Google Ads AI Features Explained: Performance Max, AI Overviews Ads, and Smart Bidding — What Advertisers Must Know

Google Ads is no longer a set of manual levers with a little automation on top. The AI inside it now decides the bid in every auction, assembles the creative, picks the placement, and even writes the ad copy — and it does all of that inside the same auction that has always decided which ads show and what they cost. This guide explains the whole AI layer as one system, in plain language: Smart Bidding sets the bid; Performance Max runs the entire campaign; AI Max optimises your Search campaigns; Demand Gen creates demand in the feeds; creative AI writes and draws the ads; and ads in AI Overviews and AI Mode put those ads inside Google’s AI-generated answers. The one-line verdict is this: the AI decides the auction, the creative and the placement; the human decides the goal, the budget, the guardrails and the message. Everything in this guide comes from Google’s own Help pages and its Ads & Commerce blog, read in full. Checked 12 November 2026. These features change fast — dates and availability in this guide are the position as of that check, and the timeline below shows how quickly the picture has moved in the last eighteen months.

The Short Answer: What Google Ads’ AI Actually Does

<img src="

” alt=”A wide desk with several small blank screens and a single large monitor showing only abstract coloured blocks and a simple upward-trending line with no numbers or labels, a person’s hands resting on the keyboard, soft evening light”>

Start with the six features in one sentence each, because the whole guide is an expansion of this paragraph. Smart Bidding is the AI that sets your bid in every single auction, choosing a bid designed to win a conversion at your target cost or to maximise the value of the conversions you get. Performance Max is a campaign type that takes your goal, your assets and your budget and runs the whole campaign across every Google surface — YouTube, Display, Search, Discover, Gmail and Maps — deciding for itself where each ad appears. AI Max for Search is not a campaign type at all: it is an optimisation layer inside your existing Search campaigns that expands which searches you match and rewrites your ad text and landing pages to fit them. Demand Gen is the campaign type for the feeds where people discover things rather than search for them — YouTube, including Shorts, Discover, Gmail, Maps and the Display Network — built to create demand rather than capture it. Creative AI is the set of tools that generates headlines, descriptions, images and videos from your landing page and your existing assets. And ads in AI Overviews and AI Mode are paid placements inside or alongside Google’s AI-generated answers to a search, labelled “Sponsored” and shown to people who are still in the middle of exploring a question.

Those six pieces are not six separate products competing for your attention. They are one layer, and they stack. A Performance Max campaign uses Smart Bidding to set its bids and creative AI to build its ads. A Search campaign with AI Max switched on uses Smart Bidding to bid and text customization to write the headlines that match the searches AI Max finds. An ad that ends up inside an AI Overview got there because a Search, Shopping or Performance Max campaign was already eligible and the AI decided the placement matched. Understanding any one feature in isolation is possible; understanding how they combine is what actually matters, because that is how the money moves.

The division of labour is the thing to hold on to through the rest of this guide. The AI is good at three jobs: bidding, matching and drafting. It can bid in an auction using signals no human could weigh one by one — the device, the time of day, the exact wording of the query, the price competitiveness of the product, the season. It can match an ad to a search the advertiser never thought of, because it understands what the search means rather than only what words it contains. And it can draft headlines, descriptions, images and videos at a volume no human team could sustain. What it cannot do is decide what your business is trying to achieve, what a conversion is worth to you, what you refuse to be associated with, or whether the words it wrote are true about your product. Those are human jobs, and they are the jobs this guide keeps returning to: the goal, the budget, the guardrails and the message.

One more thing before the detail starts. None of this replaces the auction. Every AI feature in Google Ads bids inside the same auction that has decided ad placement for years, and the auction still has the final word on what shows and what it costs. The AI does not buy its way out of the auction, and it does not set prices. It bids in it, faster and with more information than a human could. That is the foundation the next section explains, and everything after it assumes you understand it.

The pace of change is the last part of the short answer, and it is worth seeing in one place. This table lists the dated changes Google itself announced, verified from Google’s own blog and Help pages on 12 November 2026. The May 2025 row was verified for this article on Google’s own Google Marketing Live 2025 blog post, which says Google was “bringing ads to AI Mode” at that event.

Notice what the table shows: in eighteen months, a whole new ad placement appeared, brand settings moved into a new panel, bidding strategies were renamed, one campaign type began absorbing another, and a campaign type that has existed for years was given a retirement date. Anyone running Google Ads needs to re-check availability before spending, because a feature that was a test last quarter can be a default this quarter. The rest of this guide is written against the position as checked on 12 November 2026.

How the Money Works: The Auction, Ad Rank and Quality Score

Every time someone searches on Google or visits a site that shows Google ads, a lightning-fast auction runs. It runs in the time it takes the page to load, and it runs separately for each ad location on the page — the ads at the top of the search results are decided by a different auction from the ads at the bottom, and an ad can show in one location but not another. The auction is the foundation everything else in this guide sits on. The AI features do not replace it. They bid inside it.

Six factors decide which ads appear and in what order. First, your bid — but a bid in Google Ads is a maximum, not a price. You tell Google the most you are willing to pay, and what you actually end up paying is often less. Second, the quality of your ads and your landing page, summarised in a Quality Score from 1 to 10, built from three things: the expected click-through rate of the ad, how relevant the ad is to the search, and the experience of the landing page it leads to. Third, the expected impact of your ad assets and formats — the extra pieces attached to your ad, like sitelinks or images, and how likely they are to improve performance. Fourth, Ad Rank thresholds: a minimum quality an ad must reach to appear in a given position at all, which is why some searches show fewer ads than others. Fifth, the context of the ad: the search terms, the person’s location, the device, the time of day, the nature of the search terms, the other ads and results on the page, and other user signals. Sixth, the competitiveness of the auction itself: ads with similar Ad Ranks have similar chances of winning, and a wider gap makes the higher ad more likely to win — though it may pay a higher price for the privilege.

The most useful sentence in Google’s own auction documentation is this one: even if your competition has higher bids, you can still win a higher position at a lower price. That is the whole point of Quality Score and of the expected impact of your assets. A relevant ad on a good landing page can outrank a higher bid on a poor one, and the winner pays less than the loser was willing to pay. This is why the AI features matter so much and also why they cannot fix everything: Smart Bidding can bid brilliantly in every auction, but if the ad is irrelevant and the landing page is poor, the auction still punishes it. No bidding strategy repairs a bad landing page, and no amount of AI-generated copy repairs an offer nobody wants.

Quality Score deserves a warning of its own, because it is widely misunderstood. It is a summary, not a price formula. Google publishes the three ingredients — expected click-through rate, ad relevance, landing-page experience — and the 1-to-10 scale, but it publishes no formula that turns a Quality Score into a price. The actual amount you pay per click is set by the auction each time, and Google does not publish the calculation. Anyone who tells you a precise formula for what a click will cost is not quoting Google. The same caution applies to Ad Rank: Google lists the factors that feed into it and the thresholds that gate positions, but the exact arithmetic is not published.

There is one more structural fact that explains a lot of what follows. An ad shows only once in a single ad location, but it can show across locations — the same ad can appear at the top and at the bottom of the same results page, decided by two different auctions. And different auctions run for each ad location, which is why the same campaign can win the top slot for one query and lose the bottom slot for another. The AI features in this guide all operate inside this structure. Smart Bidding chooses the bid for each auction. Performance Max chooses which auctions to enter at all. AI Max finds new auctions your keywords never would have entered. The auction is the market; the AI is the trader working inside it. For readers who want the deeper explanation of what “AI” means in this context — how the underlying systems learn patterns and predict outcomes — our guide to how AI works explains transformers, neural networks and next-token prediction in plain language.

Smart Bidding: The AI That Sets Your Bid Every Auction

Smart Bidding is the oldest and most widely used piece of Google Ads’ AI, and it is also the easiest to misunderstand, because the name sounds like a single feature and it is actually four strategies. All four use Google AI to optimise for conversions or conversion value in every auction — what Google calls “auction-time bidding”. The difference between them is what you are asking the AI to optimise for, and that choice is a business decision, not a technical one.

Maximize conversions tells the AI to get as many conversions as possible within your budget, with no target cost per conversion. Target CPA — cost per acquisition — tells it to get as many conversions as possible at an average cost per conversion you set. Maximize conversion value tells it to get as much total conversion value as possible within your budget, with no target return. Target ROAS — return on ad spend — tells it to get as much conversion value as possible at an average return you set. The volume strategies fit a business that wants more leads or sales at a known cost; the value strategies fit a business whose conversions are worth different amounts — a shop where one sale is worth ten times another, for example — and that wants the AI to chase the valuable ones rather than the cheap ones. Picking the wrong one is a common and expensive mistake: a Target CPA strategy on a business whose conversions vary wildly in value will happily buy the cheap, worthless conversions and ignore the expensive, valuable ones, because you told it cost was what mattered.

What makes it “smart” is the range of signals the AI weighs in each auction. Google lists them, and the list is long: the device, the person’s physical location, their location intent, the weekday and time of day, whether they are on a remarketing list, the characteristics of the ad itself, the interface language, the browser, the operating system, the actual search query, the search network partner, the web placement, the site’s behaviour, product attributes, hotel and itinerary attributes, price competitiveness, and seasonality. No human could weigh all of that for every auction, and no human needs to — that is the job the AI is doing. But notice what the list is made of: context about the moment, not judgement about the business. The AI knows the search happened at 9pm on a phone in Lahore; it does not know whether your offer is any good.

Two practical rules govern Smart Bidding, and both are stated by Google itself. First, it needs conversion tracking. The AI learns from conversions; without conversion data it has nothing to optimise toward, and Google is explicit that conversion tracking is a prerequisite. An advertiser who switches on Smart Bidding before conversion tracking is set up correctly is asking the AI to optimise toward nothing. Second, judge it over a long enough window: Google recommends evaluating results over periods with at least 30 conversions, and 50 for Target ROAS. Fewer conversions than that and the numbers are mostly noise. This is not a suggestion to be polite about — it is the difference between concluding a strategy works and concluding it does not, on the basis of a week of data that proves neither.

One change is coming that advertisers should know about now, so it does not surprise them later. From June 2026, Google is relabelling two of the strategies: “Maximize conversions with a Target CPA” becomes simply “Target CPA”, and “Maximize conversion value with a Target ROAS” becomes “Target ROAS”. The behaviour does not change, and no action is needed. The names are catching up with how people already talk about them. Google also offers reporting tools around Smart Bidding — bid strategy reports that show the top signals driving performance, simulators that forecast cost, conversions, value, impressions and clicks at different targets, and drafts and experiments for testing a strategy against the current setup before committing. The simulator is worth using: it is the closest thing to a preview of what a different target would do, and it costs nothing to look at.

Finally, what Smart Bidding cannot do. It cannot fix a broken landing page or a wrong goal. It cannot make people want a product they do not want. It cannot tell you whether your target CPA is realistic for your market — it will simply bid as if it is, and the result will be either too few conversions or conversions that cost more than you planned. And Google makes no percentage improvement claim for Smart Bidding itself. The only percentage-type claim on the Smart Bidding help page is a narrow one about App Connect — that ad clicks landing in an app rather than a mobile website give, on average, twice the conversion rate — and that is a claim about where the click lands, not about Smart Bidding. Do not let anyone sell you Smart Bidding with a percentage improvement figure attached; Google does not publish one. For the newest development in this area, Google’s expansion of Smart Bidding Exploration to Performance Max and Shopping shows where the strategy is heading next — the AI pursuing less obvious, potentially high-performing searches beyond the ones you would have chosen yourself.

Performance Max: One Campaign, Every Google Surface

Performance Max is the campaign type that most fully hands the steering wheel to Google’s AI. It is goal-based: you tell it what you want — sales, leads, website traffic, or local store visits and promotions — and it runs a single campaign across all of Google Ads’ inventory: YouTube, Display, Search, Discover, Gmail and Maps. There is no keyword list to build and no placement list to manage. You supply four things — conversion goals and their values, creative assets in text, image and video form, audience signals that hint at who your best customers are, and a budget — and Google AI handles the rest: bidding, budget optimisation, audiences, creative assembly, attribution and placement.

The creative system inside Performance Max is built around asset groups. An asset group is a collection of creatives centred on a theme or an audience — a set of headlines, descriptions, images and videos that belong together because they speak to the same kind of customer or the same kind of product. Google AI mixes and matches the assets inside a group to render tailored ads for each placement and each person: a video-led ad on YouTube, a text-led ad on Search, an image-led ad on Discover. The advertiser does not decide which combination appears where; the AI does, and it learns from which combinations convert. That is the point of the format, and also the thing advertisers find hardest to accept: the ad your customer sees may be a combination you never previewed.

Performance Max is not a replacement for Search campaigns, and Google says so plainly: it complements keyword-based Search campaigns. The two are designed to work together, and the priority rules make that explicit. If you have an exact-match keyword in a Search campaign for the same query, the Search campaign wins — exact match is prioritised over Performance Max. Search themes, the other way of guiding Performance Max toward particular queries, carry the same priority as phrase and broad match. The practical meaning is that your Search campaigns keep the queries you care most about, and Performance Max finds the rest: the queries you never thought of, the placements you never considered, the audiences adjacent to the ones you already target. An advertiser who turns off Search and runs only Performance Max has misunderstood the design.

When to use it, in Google’s own framing, is when you have specific conversion goals, when you want to maximise performance without restricting the placements or channels the campaign can use, when you want all channels in one campaign rather than managing several, and when you want to reach customers beyond keyword-based Search. That is a wide brief, and it is deliberately wide: Performance Max is built for advertisers who would rather set a goal and a budget than manage the machinery. The budget panel reflects that philosophy. It shows up to three daily budget suggestions — Low, Medium and High — with performance forecasts attached, so you can see what Google expects each level of spend to deliver before you commit. The forecasts are estimates, not promises, and they are account-specific: the Low suggestion for one business is not the Low suggestion for another.

The honest trade-off is reach and simplicity against control and visibility. On one side, a single campaign reaches every surface Google sells, finds audiences and queries you would never have found yourself, and assembles creative combinations at a scale no human team could match. On the other side, you cannot see a placement-level cost breakdown — you do not get to know what a particular YouTube placement or a particular Discover slot cost you — and you cannot choose the placement for a given ad. The reporting tells you what happened in aggregate; it does not tell you where, in detail. Advertisers who need to know exactly where every cent went will find that frustrating. Advertisers who care about the total result at the total cost will find it liberating. Both reactions are reasonable; the mistake is expecting one campaign type to give you both.

Two asset behaviours deserve attention because they catch advertisers by surprise. First, text customization — covered in full in the creative AI section — generates additional headlines and descriptions from your landing page, domain and existing ads, and combines them with the assets you supplied. Second, if you supply no video, Google generates one automatically from your asset-group assets — and warns that a generated video may show a different product than the one on your landing page. That warning is worth reading twice. The AI is assembling creative from your materials; if your materials are ambiguous, the assembly can be wrong, and the advertiser is responsible for what it produces. The asset discussion has moved quickly in the news, too: Google now adds sourced videos and landing page images to Performance Max, with sourced videos on by default, and Performance Max will use your website photos by default — both changes that affect what the AI can pull into your ads without you uploading it. Check the defaults on your own campaigns; defaults change, and the AI uses whatever it is allowed to use.

AI Max for Search: The Optimisation Layer on Your Search Campaigns

AI Max for Search is the feature most often confused with Performance Max, so the distinction matters and it is simple: AI Max is not a campaign type. It is a continuous optimisation layer inside your existing Search campaigns. You keep the Search campaign structure you already know — ad groups, keywords, ads — and AI Max works on top of it, processing real-time signals to refine targeting and creative delivery. Performance Max replaces the campaign; AI Max upgrades it.

It has two engines. The first is search term matching. It uses broad match, asset-based and landing-page-based technology to expand the reach of your Search campaigns beyond the keywords you entered — finding searches that mean the same thing as your keywords even when they use different words, and matching your ads to the content of your landing pages. It works at ad-group level and is on by default when AI Max is on. The second engine is asset optimisation: text customization, which writes additional headlines and descriptions from your landing page, domain and existing ads, and Final URL expansion, which can replace the final URL of your ad with a more relevant landing page from your site and generate matching dynamic copy for it. Both are turned on when AI Max is on, and both can be toggled individually — except that Final URL expansion requires text customization to be switched on first.

AI Max also folds in the settings that used to live in separate places. Brand settings — inclusions and exclusions — live in the AI Max panel now: inclusions at campaign and ad-group level, exclusions at campaign level. Locations of interest, URL inclusions and exclusions are part of the same panel. That consolidation is the reason the timeline in the first section matters: brand settings for Search campaigns began upgrading into AI Max on 27 May 2025, and advertisers who set brand exclusions years ago should check that they transferred as expected rather than assuming they did.

The reporting AI Max adds is one of its strongest arguments, because it answers the question Performance Max leaves open: what did the AI actually do? The search terms report gains an “AI Max” match type and a source column that distinguishes broad-match expansion from keywordless matching, so you can see which of your impressions came from a keyword you chose and which came from the AI’s own matching. There is a combined view of search term, headline and URL, a keywords report with summary rows, a landing pages report with a “Selected by” column showing whether the landing page was yours or the AI’s choice, and an asset report. This is meaningfully more transparent than Performance Max, and it is the reason advertisers who want AI assistance without giving up visibility often prefer AI Max: you can see the AI’s work, section by section, and intervene where you disagree with it.

There are two warnings attached to AI Max that are easy to miss and expensive to ignore. The first is the tracking-template warning: dynamic landing pages can return 404 errors if your tracking template lacks a compatible {lpurl} tag, because Final URL expansion sends people to pages your template was not built for. Check the template before switching it on, not after the clicks start failing. The second is the budget warning: AI Max will not be effective if the campaign is limited by budget, and Google shows an alert when that is the case. An AI that cannot bid freely because the budget ran out at noon is not an AI problem; it is a budget problem, and the fix is money, not settings.

Two more facts complete the picture. New Search campaigns have AI Max on by default, so a new advertiser is using it whether they know it or not — check your campaign settings if you have not looked since you created the campaign. And from February 2027, Dynamic Search Ads campaigns will automatically upgrade to AI Max; DSA was the older way of doing keywordless Search, and AI Max is its successor. The difference between the two features, in Google’s own words, is that AI Max optimises Search campaigns while keeping the familiar Search structures, where Performance Max optimises across all channels from a single campaign. The choice between them is a choice about how much structure you want to keep.

Demand Gen: AI for the Feeds Where People Discover

Demand Gen is the campaign type for the other half of the internet: the places where people are not searching for something but discovering it. It serves YouTube, including Shorts, Discover, Gmail, Maps and the Google Display Network — the visual, feed-based surfaces where a person scrolls, watches and browses. Its job is demand creation, not demand capture. A Search campaign catches someone who already knows what they want and is looking for it. A Demand Gen campaign puts a product in front of someone who did not know they wanted it yet, and tries to make them want it. That is a different job, and it needs different creative: visual, entertaining, built for a feed rather than a results page.

The scale of those surfaces is the reason Demand Gen exists as a separate campaign type. Google states that the surfaces it reaches have up to 3 billion monthly active users, and that Shorts alone gets more than 50 billion global daily views. Those are Google’s own figures, and they explain why a campaign built for feeds needs its own tools rather than borrowing Search’s. The AI combines visuals, messages and placements — choosing which image or video to show, which message to pair it with, and which surface to show it on — and the advertiser supplies the raw material and the goal.

Demand Gen gives the advertiser more control than Performance Max does, and the difference is deliberate. It offers Lookalike segments built from past purchasers, site visitors and YouTube viewers, so you can ask for more people like the people who already buy from you. It offers creative A/B experiments, so you can test one version of an ad against another and learn which one works. It supports product feeds and promotion assets, so a retailer can show actual products with prices rather than only brand imagery. It lets you choose the channel — you can serve on Shorts only, for example, rather than everywhere at once. And its bidding options are clicks, conversions or conversion value, with value-based bidding and Maximize Clicks available. The comparison Google itself draws is the clearest summary: Search maximises Search performance, Performance Max maximises across all channels, and Demand Gen generates demand on the most visual, entertaining surfaces, with more control over creative testing, audience targeting and specific placements.

Measurement in Demand Gen is built around lift: brand lift, search lift and conversion lift — whether the campaign moved the needle on awareness, on searches for your brand, and on sales, rather than only on clicks. That fits the job: a feed campaign’s value often shows up later, as a search or a purchase, rather than in the click that follows the ad. The budget guidance is specific and worth noting: for target-CPA campaigns, Google’s help page recommends a budget of at least ten times the target CPA. That is one of the very few concrete budget numbers Google publishes anywhere in its Ads documentation, and it is a useful rule of thumb for sizing a Demand Gen test: a target CPA with a budget that cannot afford ten conversions is a budget that cannot teach the AI anything.

One structural change is under way and advertisers with Display campaigns need to know about it. Display campaigns are moving to Demand Gen: voluntary migration began in June 2026, new campaigns will only be creatable in Demand Gen later, and remaining eligible campaigns will be auto-migrated after that. If you still run Display campaigns, the migration tool is already available, and the eventual auto-migration means the decision is when, not whether. There are also limits worth knowing: sensitive categories are generally restricted to YouTube only, not Gmail or the Display Network, and alcohol campaigns may not use product feeds. Brand safety is handled at account level through content suitability settings, with placement exclusions available through manager accounts and keyword exclusions at account level. The feed is a powerful place to advertise, and it is also a place where the wrong adjacency — your ad next to the wrong content — does real damage. The controls exist; use them before the campaign goes live, not after someone screenshots it.

Ads Inside AI Overviews and AI Mode: The Newest Placement

The newest placement in Google Ads is also the strangest, because the ad appears inside an answer rather than beside a list of links. When someone asks Google a question and gets an AI Overview — a generated summary at the top of the results — or opens AI Mode — the conversational search experience — ads can now appear within or alongside the AI-generated response. They are labelled “Sponsored” and visually distinguished from the organic content, and Google is explicit about the boundary that matters most: paid advertising does not influence the organic AI responses. Paying for an ad does not change what the AI says, and being cited in the organic answer does not affect whether your ad is eligible. The two systems sit in the same answer and do not touch each other.

The scale of the change is recent. Ads in AI Mode were announced for the US at Google Marketing Live on 21 May 2025. On 28 October 2026, Google expanded them to ten more countries — Australia, Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines and Singapore — and launched multiturn ads in the US, so ads can appear in the follow-up questions of a conversation, not only in the first answer. Expansion to additional locales, including Brazil, Mexico and the United Kingdom, is planned for Q4 2026 — planned, not live, as of the 12 November 2026 check. Google’s own research, quoted in the announcement, says people are comfortable with ads in AI Mode and find them highly relevant, and that AI Mode queries are, on average, three times longer than traditional searches. That last claim carries footnotes whose detail was not visible in the read for this guide, so the scope and sample behind it are not stated here. Google also says click-through rates are already similar between sponsored and organic product list items within AI Mode — but that claim is specifically about product list items, not about text ads, and no cost-per-click or conversion data was given. Do not stretch either claim further than Google stated it.

As of the check for this guide, the in-answer placements run in English only, in twelve markets. The list, from Google’s own help page, is: the United States, Australia, Canada, India, Indonesia (AI Overviews only), Kenya, Malaysia, New Zealand, Nigeria, Pakistan (text ads only), the Philippines and Singapore. Ads above or below an AI Overview are a different matter — those run in all 200-plus countries and territories and all languages where AI Overviews are available. The distinction matters: the ad above the answer is the old placement in a new position, and it is everywhere; the ad inside the answer is the new placement, and it is in twelve English-language markets. Google notes the list may not be exhaustive, so check the help page before planning spend around a market that is not on it.

There is nothing to set up, and that is the point. Existing Search, Shopping and Performance Max campaigns are automatically eligible. There is no new campaign to create, no separate budget to set, and no extra charge beyond the normal auction — the same cost-per-click or conversion-based bidding, with the standard invalid-traffic protections. Text ads and Shopping ads can appear; local inventory ads and vehicle ads are being experimented with, which means they are tests, not settled placements. Matching works in two ways. Direct matching is the familiar kind: the query itself shows clear commercial intent, and the ad matches the query. Contextual matching is the new kind: the intent emerges from the AI response rather than the query — Google’s own example is a “why is my pool green” question surfacing ads for pool vacuums, where the query is a problem and the response is where the commercial intent appears. AI-powered targeting — AI Max, broad match, Shopping and Performance Max — is eligible for both matching types. Exact and phrase match are eligible only for direct commercial intent, because a keyword that demands an exact wording cannot match an intent that only appears in the generated answer.

One feature of these placements surprises advertisers more than any other, and it should be understood before a campaign runs rather than after. Some placements carry AI explainers: short text generated by Gemini explaining how an ad relates to the search. Advertisers cannot write them, edit them, approve them or opt out of them. They are distinct from ad copy, and they are generated by Google’s model, not by the advertiser. If the idea of Gemini writing a sentence about your product that you cannot review bothers you, that is a reasonable reaction — and it is also the current position, stated plainly in Google’s help page. The same plainness applies to the three things advertisers cannot do in these placements at all. You cannot target only AI Overviews or AI Mode; there is no switch for that. You cannot opt out of these placements separately from Search — opting out means opting out of Search entirely, though negative keywords and brand exclusions still apply. And you cannot see per-placement reporting: the metrics are included only in aggregate in standard campaign and search-terms reports. You will know the placements contributed; you will not know exactly how much.

Google’s own best practices for these placements are four, and they are a useful summary of the whole guide: use AI-powered targeting, because exact match alone cannot reach the contextual matches; use Smart Bidding with first-party conversion data and flexible budgets, so the AI can bid effectively in auctions it has never seen before; use text customization and Final URL expansion, so the ad and the landing page fit the intent the AI found; and keep Merchant Center feeds complete and fresh, so Shopping ads have accurate products to show. The placement is still evolving — layouts are continuously tested and may change, and an ad can appear both above and within an AI Overview on the same search, though not both within and immediately below it. The site’s own reporting captures that evolution as it happens: Google’s AI Mode now stacks four ads under one answer is a live example of a layout test in progress. For the reader’s side of the same experience — what AI Overviews and AI Mode look like to the person searching, and how to keep the classic results alongside them — our guide to Google’s AI search features covers it. And for the organic side of appearing inside an AI answer, which is a different system from the paid side entirely, our GEO guide explains how content gets cited by AI answers — remember that a paid ad does not influence the organic answer, and an organic citation does not affect ad eligibility.

Creative AI: Text Customization, Asset Generation and Asset Studio

Creative AI is the part of Google Ads’ AI that most directly touches what the customer sees, and it is the part where the advertiser’s responsibility is stated most bluntly. There are three tools in the set, and they do three different jobs: text customization writes words, generative image tools draw pictures, and auto-generated video makes moving pictures out of still ones. All three work from the same raw material — your landing page, your domain, and the ads you already have — and all three require the same thing from you afterwards: a human review before anything goes live.

Text customization, formerly called “automatically created assets”, generates additional headlines and descriptions from your landing page, your domain and your existing ads, and combines them with the assets you supplied yourself. It is the engine inside AI Max’s asset optimisation and inside Performance Max’s creative assembly, and it can also be used on its own. Final URL expansion is its partner: it can replace the final URL of your ad with a more relevant landing page from your site and generate matching dynamic copy for the new destination. The two together mean that the ad a customer sees — its words and its destination — may both have been chosen by the AI rather than by you. Google states the responsibility plainly: the advertiser remains responsible for reviewing landing-page content and all dynamically generated assets. That is not a formality. The AI writes from your site, and if your site contains an outdated price, a discontinued product or a claim you no longer stand behind, the AI can put it in an ad. The review step is the only thing standing between your website’s stale corners and your customers.

Auto-generated video works the same way, with one warning that deserves emphasis. When no video is supplied in an asset group, Google creates one automatically from the asset-group assets — and warns that a generated video may show a different product than the one on the landing page. A video that shows the wrong product is worse than no video at all: it is an ad that misleads the person who clicks it. If you run Performance Max or any campaign where video can be generated, check what the AI made before the campaign spends. The check costs minutes; the wrong video costs the campaign’s credibility.

Generative image tools create new image assets in seconds from a prompt, tailored to the business and brand, or generated automatically from your text assets or landing page. They are found in campaign setup, in ad and asset-group editing, in the Asset Library, and in certain recommendations. The more interesting capability is product-featured generation: you provide your own product images, and the tool builds lifestyle imagery around them — people interacting with or wearing the products, in different orientations and settings. Style-reference images, up to five per generation, let you steer the look, though they must not contain logos, watermarks or product images. There are eligibility limits: manually prompted generation requires an interface in English, Dutch, French, German, Italian, Japanese, Portuguese or Spanish; an account that has been active and running ads for at least 30 days; a history of policy compliance; no completed Election Advertising Verification; and an account outside a sensitive vertical. The rollout is gradual, so an eligible account may still not see the tools yet. Do not promise a client that image generation is available until you have opened their account and seen it.

Two rules govern what happens to generated images after they are made. First, they stay in the Asset Library for 14 days unless they are used in a campaign — after that they disappear, so an image you like needs to be put to work or it is gone. Second, all generated images carry identification mechanisms, including SynthID, an imperceptible digital watermark that survives screenshots, filters and compression and can be checked at Google’s SynthID tool. That watermark is part of a wider disclosure regime: regulations in the EU, India, New York, California, South Korea and Vietnam require disclosures or labels on certain AI-generated or edited ad assets. Google has expanded its AI-label coverage as those rules have spread — our report on the AI-label expansion to California, Korea and Vietnam tracks the latest additions — but Google also warns that its AI-label setting does not guarantee compliance. The label is a tool; the legal responsibility is still yours, and it varies by market.

The safety limits are worth stating because they define what the tools will not do. Google will not automatically suggest generated content for sensitive verticals such as political or pharmaceutical advertising. People and face generation is limited to adults — no children, no prominent likeness. Branded items or logos that were not provided as input are not generated. And the tools are experimental: Google says generative AI “may display inaccurate or offensive content”, which is why the review rule exists. The advertiser must review generated assets for accuracy and policy compliance before publishing. Every generated image, headline and video is a draft until a human has read it. Treat it as one.

The Conversational Layer: Ask Advisor and AI Campaign Help

Ask Advisor is the most recent addition to the AI layer, and it changes the shape of the work rather than the work itself. It is an agentic conversational experience built with Gemini, inside Google Ads, currently in beta. Instead of navigating menus and reports, you ask it questions in plain language — why did performance change this week, what is wrong with this ad, why was this campaign disapproved — and it analyses the account, answers, troubleshoots performance or ads-policy issues, and suggests new text and creatives. It is surfaced inside Performance Max and other campaign workflows, which is to say it appears where the work is already happening rather than in a separate tool you have to remember to open.

The honest limit is the same one that applies to every assistant: it suggests, it does not decide. Ask Advisor can tell you that conversions fell because a landing page started returning errors, or that a campaign is limited by budget, or that a headline violates a policy — and those are genuinely useful answers that used to require digging through reports. But every suggestion it makes still needs a human yes before it changes anything in the account, and the human yes is the part that matters. A beta assistant that analyses well is a good tool. A beta assistant treated as a manager is a mistake. The distinction is the same as the one this guide has drawn throughout: the AI does the analysis and the drafting, and the strategy and the accountability stay human.

Ask Advisor is also the clearest sign of where Google Ads is heading. The interface is becoming a conversation: you describe the outcome you want, the AI proposes the campaign, the creative and the settings, and you approve or correct it. The campaign-translation tool is a smaller example of the same direction — AI doing the repetitive work that used to eat an afternoon. Google’s tool that translates entire campaigns takes a campaign built for one language and rebuilds it for another, which is exactly the kind of mechanical, high-volume work that AI does well and humans do slowly and inconsistently. Expect more of these: the repetitive work disappearing into the AI, and the judgement work — what to say, who to say it to, what it is worth — staying with the advertiser. The advertisers who do well in that world will be the ones who are good at the judgement work, because the mechanical work is becoming free.

Brand Controls: What You Can Still Fence Off

For all the AI’s freedom to choose placements, audiences and creative, there is a set of controls that still fence it off — and knowing them is the difference between an advertiser who uses the AI and an advertiser who is used by it. The controls are the subject of this section, and the theme is simple: every control narrows reach, so each one should be used where it genuinely matters and nowhere else. Google itself warns advertisers to use brand settings only where really needed, because a fence that is too tight starves the AI of the room it needs to find conversions.

Brand inclusions and brand exclusions are the pair most people confuse, and they do opposite jobs. A brand inclusion, available in Search only, tells a campaign to serve only for the brands you select — it is a whitelist, and it narrows the campaign to those brands alone. A brand exclusion tells a campaign it will not serve for the brands you select — it is a blacklist, and it keeps your ads away from searches for those brands. Exclusions apply in Search and in Performance Max; in Performance Max they apply to Search, Shopping and YouTube search inventory, with an option to still allow Shopping ads on excluded-brand searches. The practical uses are obvious once you see them: exclude your own brand from a Performance Max campaign so it does not spend your prospecting budget on people who were already searching for you by name, or exclude a competitor’s brand so your ads never appear on their searches. The two controls moved house in 2025: Search brand settings began upgrading into AI Max on 27 May 2025, so they now live in the AI Max panel rather than where long-time advertisers remember them.

Negative keywords are the next fence, and they have two levels. At campaign level, they stop an ad showing for a particular search term. At account level, you can block up to 1,000 negative keywords across the whole account. In Performance Max, negative keywords apply to Search and Shopping inventory only — they do not reach the YouTube, Display, Discover, Gmail or Maps placements, because there is no search query to match against in a feed. That limit surprises people, and it is worth knowing before you rely on a negative keyword to keep an ad out of a placement it cannot reach. Placement exclusions are the control for those surfaces: you can exclude specific placements — a website, an app, a YouTube channel — from showing your ads. Content suitability settings, set at account level, control the kind of content your ads appear alongside: the inventory type, the content types and the labels. Content-theme exclusions, also at account level, exclude whole themes. Manager accounts can apply placement exclusions across accounts they manage.

The softer controls are guidance rather than fences, and they are often more useful than the hard ones. Search themes tell Performance Max which queries matter to you — they carry the same priority as phrase and broad match keywords, which means they are a strong steer rather than a command. Keywordless targeting, the other side of the same coin, lets the AI find searches from your landing pages rather than your keyword list. Channel prioritization lets you set priority tiers across inventory — Google channels grouped together, with Search partners and the Display Network as separate tiers, and optional daily budget allocations per tier — so you can say, in effect, that Search matters more to you than the Display Network without switching either off. And asset pinning lets you fix a headline or description to a position in the ad, so a legally required line or a brand message always appears where you put it — with one exception worth knowing: pinned assets are not used when Final URL expansion picks a more relevant landing page, because the pinned copy may not fit the new destination.

One more control is easy to forget and matters more than it looks: the URL inclusions and exclusions inside AI Max. URL inclusions tell the AI which pages it may send people to; URL exclusions tell it which pages it may not. Combined with Final URL expansion, they are the difference between an AI that sends traffic to your best pages and an AI that sends traffic to a page you forgot existed — a careers page, a terms page, an old product page with a price you no longer charge. Set them when you switch AI Max on, not after the first surprising click. The same logic applies to every control in this section: the AI is powerful precisely because it explores, and exploring without fences is how an ad ends up in a place you would never have chosen. Fences do not make the AI worse. They make it yours.

What the AI Can Actually See: Insights, Reporting and What Stays Hidden

The AI layer produces a flood of data, and the skill is knowing which of it to read and which to ignore. Google’s reporting is built around two ideas: the Insights page, which tells you what changed and why, and the campaign and asset reports, which tell you what happened in detail. Both are useful. Neither tells you everything, and the gaps are as important as the contents — because the two hardest limits in this guide live in the reporting, not in the features.

The Insights page refreshes daily, and it is the first place to look when something moves. Search terms insights analyse the past 28 days, grouped into categories, with metrics, for Performance Max, Search and Shopping campaigns — so instead of a list of thousands of individual queries, you see the categories of searches that are driving your account, and which of them are growing. Audience insights show who is responding, including asset audience insights that connect particular assets to particular audiences. Performance insights explain significant changes in the past 7 or 28 days and their root causes — a conversion drop is not just reported, it is explained, with the likely cause named. Demand forecasts flag categories with estimated significant growth in the coming months, so you can prepare a campaign before the demand arrives rather than after it. The details in this paragraph are reported from the Insights page’s documentation as summarised in the research for this guide; the specifics were not re-verified in full, so treat the shape as right and check the page itself for the current layout.

Below the Insights page sit the reports that answer specific questions. Bid strategy reports show the top signals driving a strategy’s performance — the device, the location, the day and time, the queries, the Customer Match lists that mattered most — which is the closest Google comes to showing you why Smart Bidding bid the way it did. Asset reporting rates individual assets, so you can see which headlines, images and videos are working and which are dragging the group down; asset-group reporting does the same at group level. Channel performance reports show how the campaign performed across the channels it reached. And for AI Max, the search terms report adds the “AI Max” match type and a source column, the keywords report adds summary rows, and the landing pages report adds a “Selected by” column — the reporting that lets you audit the AI’s choices rather than merely trusting them.

Then the two hard limits. First, AI Overviews and AI Mode metrics are aggregate-only: they appear inside standard campaign and search-terms reports, but you cannot break them out and see what the AI placements did on their own. You will know the placements are contributing; you will not know by how much, at what cost, or at what conversion rate. Second, Performance Max gives no placement-level cost breakdown: you cannot see what a particular YouTube placement or Discover slot cost you. These are not oversights. They are structural choices about what Google reports, and they are unlikely to change soon — Google’s own channel performance reporting for Performance Max is the closest it comes, and it is still not a placement-level cost table. Plan around them: judge Performance Max and the AI placements on their total result, not on a per-placement breakdown you will never get.

Two habits make the reporting useful rather than noisy. The first is the evaluation rule from the Smart Bidding section: judge a strategy over periods with at least 30 conversions, 50 for Target ROAS, and wait at least 2 weeks after enabling AI Max before making changes such as adding negative keywords. The AI needs time to learn, and changing the settings every few days resets the learning each time. The second is to act on categories, not single queries. A single odd search term in the search terms report is noise; a whole category of searches you never expected is a signal — about your market, about your landing pages, or about a gap in your negative keywords. The Insights page is built for exactly that kind of reading: categories, trends and root causes, not a list of one-off curiosities. Read it weekly, act on what repeats, and ignore what appears once.

What It Costs: Budgets, Billing and Why There Is No Price List

The question every advertiser asks first — what will it cost? — is the one Google answers least directly, and the reason is structural rather than secretive. There is no fixed price list for Google Ads, because there is no fixed price. Every click’s price is set by a fresh auction, and the auction is different every time: different competitors, different context, different Quality Scores. The bid you set is a ceiling, not a price, and what you actually pay is often less than that ceiling. That is the whole pricing model, and it is why no one — not Google, not an agency, not a blog — can tell you in advance what a click will cost. Anyone who does is guessing, and the guess is usually wrong.

What you do control is the budget, and the budget has hard edges that Google states plainly. You set an average daily budget per campaign. For most campaigns, your daily spend will not exceed twice that average daily budget on a given day, and your monthly spend will not exceed 30.4 times it — 30.4 being the average number of days in a month. Those are caps, not targets: a campaign can spend less than its budget on a quiet day and more on a busy one, as long as it stays inside the monthly cap. The 2-times daily allowance exists because demand is uneven — some days are worth more than others, and Google spends more on the days that are worth more. If the idea of a campaign spending double its daily budget on one day alarms you, remember the monthly cap: the total for the month is bounded, and the daily swing is how the AI buys the good days.

What the bid strategy changes is how you are billed and what the AI is optimising for. With manual or cost-per-click bidding, you pay per click and the bid is the lever. With Smart Bidding, you pay per click too, but the bid is set by the AI in each auction toward a conversion goal — a target cost per acquisition, a target return on ad spend, or simply as many conversions or as much conversion value as the budget allows. The billing model does not change; the decision behind the bid does. That distinction matters when you read the invoice: the line items look the same, but the reason a particular click cost what it did is a decision the AI made in a fraction of a second, based on signals you never saw.

The budget panels in Performance Max and Demand Gen are the closest Google comes to forecasting cost, and they work the same way in both: up to three daily budget suggestions — Low, Medium and High — with performance forecasts attached. The suggestions are account-specific, built from your own history and your market, which means they are not transferable: the Medium suggestion for one advertiser is meaningless for another. Demand Gen adds one more concrete number, and it is one of the few Google publishes: for target-CPA campaigns, budget at least ten times the target CPA. A campaign whose budget cannot afford ten conversions at its target cost cannot generate enough data for the AI to learn from, and it will spend its budget on the first few auctions of the day and then go quiet.

That brings us to the figures you will find everywhere else on the internet and should not trust. Third-party “average CPC” figures — the kind published by marketing blogs and benchmarking companies — are not Google’s figures, and they must not be quoted as if they were. They are averages of self-reported or sampled data, from different markets, different industries, different years and different campaign types, and they tell you nothing about what your click will cost tomorrow. Google publishes no average cost per click, no typical cost per acquisition, and no guaranteed return, for any of these features. This guide states no price, no average and no guarantee anywhere in it, because there is none to state. If a consultant, a course or a blog quotes you an average CPC for your industry and presents it as a fact about Google Ads, ask where the number came from — and if the answer is not Google, treat it as a rumour, not a rate card.

The practical budget advice Google does give is simple and worth following. Give high-performing campaigns flexible budgets, so the AI can bid effectively when the good auctions appear rather than running out of money at noon. Do not starve AI Max with a budget-limited campaign, because a campaign that cannot bid cannot learn. And size the budget to the learning period: a strategy judged over 30 conversions needs a budget that can buy 30 conversions, and a Demand Gen test sized at ten times its target CPA needs a budget that can afford that. More money is not more intelligence — a larger budget does not make the AI smarter — but too little money is a real constraint, because an AI that cannot bid is an AI that cannot learn. The budget is not the strategy. It is the fuel the strategy runs on.

Mistakes and Myths

Most of the money wasted on Google Ads’ AI features is not wasted on the features themselves. It is wasted on beliefs about them that are wrong. This section takes the eight that cost advertisers the most, and corrects each one with what Google itself says.

“The AI will find the right audience on its own, so I don’t need conversion tracking.” Wrong, and it is the most expensive myth on this list. Smart Bidding needs conversion tracking — Google states it as a prerequisite, not a recommendation. The AI learns from conversions; without conversion data it has nothing to optimise toward, and it will bid as if every click were equally valuable. AI Max needs conversion data to work too. Setting up conversion tracking is not admin. It is the step that makes everything else in this guide possible, and it comes before any AI switch is touched.

“Performance Max replaces Search campaigns.” Google says the opposite: Performance Max complements keyword-based Search campaigns. The priority rules prove it — an exact-match keyword in a Search campaign wins over Performance Max for the same query, by design. The two are built to work together, with Search keeping the queries you care most about and Performance Max finding the rest. An advertiser who turns off Search and runs only Performance Max has removed the one campaign type that guarantees their most important queries are covered.

“I can opt out of AI Overviews ads.” No. There is no opt-out for AI Overviews or AI Mode ads separately from Search. The only way out of those placements is out of Search entirely, and that means giving up the largest source of intent-driven traffic Google has. What remains available is the ordinary control set: negative keywords and brand exclusions still apply to the AI placements, so you can keep your ads away from specific queries and brands. But the placements themselves are part of Search now, and they are not separately switchable.

“My ad will appear in AI Mode if I just turn something on.” There is nothing to turn on, and automatic eligibility is not a promise of an impression. Existing Search, Shopping and Performance Max campaigns are automatically eligible for the AI placements, but eligibility means the ad can appear — not that it will. The auction still decides, the matching still has to find intent, and the placement is in English only, in twelve markets. An advertiser outside those markets, or in a language other than English, will not see their ads in AI answers no matter what they switch on. And a related confusion: Bing’s AI search is a different ecosystem with its own ad system — Google’s AI placements and Microsoft’s are not the same inventory, and an ad eligible in one is not eligible in the other.

“AI-generated copy is ready to publish.” Google says the opposite, twice over: the advertiser is responsible for reviewing landing-page content and all dynamically generated assets, and generative AI “may display inaccurate or offensive content”. Text customization writes from your landing page, and if your landing page is out of date, the copy it writes will be out of date. Auto-generated video can show a different product from the one on the landing page. Generated copy is a draft. It becomes an ad when a human has read it and approved it — not before.

“A higher budget guarantees better AI results.” More money is not more intelligence. A larger budget gives the AI more auctions to bid in, which helps it learn faster — but it does not make the AI smarter, and it does not fix a bad goal, a broken landing page or a wrong target. What a budget does do is remove a constraint: a budget-limited campaign blunts AI Max, because an AI that cannot bid cannot learn. The correct way to think about budget is as fuel, not as quality. Enough fuel for the learning period — 30 conversions for Smart Bidding, two weeks for AI Max, ten times the target CPA for Demand Gen — and then judge the result. Pouring more fuel into a campaign with the wrong goal just burns money faster.

“The average CPC I read online is what I’ll pay.” Google publishes no average cost per click, no typical cost per acquisition and no guaranteed return. The price of a click is set by the auction each time, and the auction is different every time. A third-party benchmark is an average of someone else’s data, from a different market and a different year, and it tells you nothing about your auction tomorrow. The same goes for the performance percentages that circulate about Smart Bidding: Google makes no percentage improvement claim for Smart Bidding itself. The only percentage-type claim on its help page is the narrow App Connect figure about app versus mobile-web landing, and that is a claim about where the click lands, not about the bidding. If a number is not from Google, do not treat it as Google’s.

“Performance Max shows me everything.” It does not, and the reporting gap is structural. Performance Max gives no placement-level cost breakdown, and AI Overviews and AI Mode metrics are aggregate-only. You will know the total result and the total cost, but you will not know what a particular placement cost or what the AI placements contributed on their own. Advertisers who plan their reporting around per-placement detail will be disappointed; advertisers who judge on total return will be fine. Know which one you are before you build the campaign.

Your Control Checklist: The Human Jobs AI Does Not Do

Everything in this guide reduces to seven jobs, and they are all human. Print this section, work through it before any AI switch is touched, and revisit it whenever a campaign is changed. The features will keep moving — the checklist will not.

1. Set the goal and the conversion value before touching any AI switch. Decide what a conversion is and what it is worth to you, in money, before you open a campaign. The AI optimises toward the goal you give it; a vague goal produces vague results, and a goal worth nothing to your business produces conversions worth nothing to your business. Value-based bidding only works if the values are real.

2. Fix conversion tracking first. Google tag, offline conversion imports, whatever your business needs — get it working and verify it before Smart Bidding, AI Max or Performance Max is switched on. The AI learns from conversions, and a campaign that starts without tracking is a campaign that starts blind. Check the tracking is recording the conversions you care about, not just the ones that are easy to count.

3. Feed the AI good assets — and review everything it generates. Text, images, video, product feed: the AI assembles what you give it, and it cannot make good ads from bad material. Then read what it made. Check the headlines text customization wrote, the video it generated, the images it produced, the landing pages Final URL expansion chose. The review step is the advertiser’s responsibility, stated by Google, and it is the last line of defence between your site’s stale corners and your customers.

4. Set the guardrails that matter — and nothing more. Brand exclusions where your own brand would eat the prospecting budget. Negative keywords for the searches you never want. Placement and content exclusions where adjacency would damage you. URL exclusions for the pages the AI must never send people to. Then stop. Every control narrows reach, and Google warns to use brand settings only where really needed. A fence around everything is a campaign with nowhere to go.

5. Give it a real budget and a real evaluation window. Not a budget-limited campaign — AI Max will not work properly in one — and not a budget that cannot afford the learning period. Thirty conversions before you judge Smart Bidding, fifty for Target ROAS. Two weeks after enabling AI Max before you change anything, including adding negative keywords. Ten times the target CPA for a Demand Gen test. The AI needs time and data to learn, and changing the settings every few days resets the learning each time.

6. Read the Insights page and the search terms report weekly — and act on categories, not single queries. The Insights page refreshes daily and explains significant changes with their root causes. The search terms report shows which categories of searches are driving the account. A single odd query is noise; a whole category you never expected is a signal about your market, your landing pages or a gap in your negative keywords. Act on what repeats. Ignore what appears once.

7. Re-check availability before spending. AI Mode ads are still English-only in twelve markets, and the expansion to Brazil, Mexico and the United Kingdom is planned for Q4 2026, not live as of this check. Image generation is rolling out gradually and may not be available in your account yet even if you are otherwise eligible. Multiturn ads are in the US only. Features that were tests last quarter are defaults this quarter, and the timeline in the first section of this guide shows how fast that happens. Check the help page for the feature you are about to spend on, on the day you spend on it.

The AI inside Google Ads is a bidding, matching and drafting engine. It bids in every auction with more information than any human could weigh. It matches ads to searches no keyword list would ever have found. It drafts headlines, images and videos at a volume no team could sustain. None of that is strategy. The strategy — what the business is trying to achieve, what a customer is worth, what the brand will and will not be seen next to, whether the words in the ad are true — is still human, and it always will be. The advertisers who do best with these tools will not be the ones who hand over the most control. They will be the ones who are clearest about what they are asking the AI to do, and who check its work every week. The AI decides the auction, the creative and the placement. You decide the goal, the budget, the guardrails and the message. Keep it that way.

Sources

All sources below are Google’s own Help pages or Google’s Ads & Commerce blog, read in full on 12 November 2026 unless noted. Checked 12 November 2026.

• Google Ads Help, “About Performance Max campaigns” — https://support.google.com/google-ads/answer/10724817
• Google Ads Help, “About Smart Bidding” — https://support.google.com/google-ads/answer/7065882
• Google Ads Help, “About ads in AI-powered experiences in Google Search” — https://support.google.com/google-ads/answer/16297775
• Google Ads & Commerce Blog, “Ads in AI Mode expand to 10 new countries” (28 Oct 2026) — https://blog.google/products/ads-commerce/ads-in-ai-mode-global/
• Google Ads Help, “How AI Max for Search campaigns works” — https://support.google.com/google-ads/answer/15910187
• Google Ads Help, “About Demand Gen campaigns” — https://support.google.com/google-ads/answer/13695777
• Google Ads Help, “About generated images in Google Ads” — https://support.google.com/google-ads/answer/15701328
• Google Ads Help, “How the Google Ads auction works” — https://support.google.com/google-ads/answer/6366577
• Google Ads Help, “About the Insights page” — https://support.google.com/google-ads/answer/10256472 (specifics reported from the search excerpt, not re-verified in full — see the research brief)