Ads in AI Search: Google AI Overviews, AI Mode and ChatGPT Ads Explained

Ads in AI search are paid placements shown inside or next to AI-generated answers. Google has announced ads in AI Overviews and ad formats in AI Mode, and OpenAI has begun testing ads in ChatGPT. For advertisers the practical message is the same: your existing Search, Shopping and Performance Max setups may become eligible, so conversion tracking, landing pages and product data matter more, and you should test with caution and judge on real results.

For years, search ads sat above and beside a list of links. As answers move to AI-written summaries and conversations, platforms are building ad formats into those experiences. This is changing quickly, and reliable detail is limited, so this guide separates what the platforms have announced from what is reported or unknown. It explains where ads are appearing, who is eligible, what controls advertisers have, how to prepare, how to measure and what it means for organic visibility. Check each platform's current documentation, because announcements and availability by country change often. For PPC basics, see our PPC advertising guide and the AI layer in AI in PPC.

What has been announced

Platform	What is announced or reported	Status caution
Google AI Overviews	At Google Marketing Live 2025, Google said it was expanding Search and Shopping ads shown within AI Overviews, extending to desktop in the US and to English in select countries later	Availability varies by country and over time
Google AI Mode	Google said it was testing new ad formats integrated into and below AI Mode responses	Described as a test at the time of the announcement
Google 2026 announcements	At Google Marketing Live 2026 (20 May 2026), Google described reinventing ads for AI search, continued AI Max for Search, a unified advisor agent across its ad and analytics products and agentic commerce work including the Universal Commerce Protocol	Summary only. Check Google's pages for rollout details
ChatGPT	OpenAI has begun testing ads in ChatGPT, widely reported as starting in February 2026 for logged-in adult users on certain plans in the United States, shown as labelled sponsored units separate from the answer	Reported details come from secondary sources. Check OpenAI's own pages
OpenAI crawler for ads	OpenAI's crawler documentation lists OAI-AdsBot, which validates the safety of pages submitted as ads on ChatGPT and, per OpenAI, may use landing page content to decide when an ad is most relevant	Confirmed in OpenAI documentation

Other AI products have also experimented with sponsored content. We do not cover them here because we could not verify current details. Treat any list of ad products as perishable.

How Google says advertisers become eligible

According to Google's Marketing Live 2025 announcements, advertisers already using Performance Max, Shopping and Search campaigns with broad match, including AI Max for Search, are eligible to have their ads appear in AI Overviews and AI Mode, with no separate campaign needed. The practical point is that visibility in these placements depends on how your existing campaigns are set up, including broad match settings, creative assets and conversion goals. Eligibility, controls and reporting detail may change, so read the current Google Ads help pages.

What the ads look like

Platforms describe the ads as clearly labelled and visually separated, but formats vary.

• Inside or below AI Overviews and AI Mode answers. Search or Shopping ads relevant to the query, shown with a "Sponsored" label, positioned within or alongside the generated response.
• In ChatGPT. As reported, sponsored cards below an answer, with a headline, short description, image or logo and a link, labelled as sponsored and kept separate from the answer. OpenAI has said ads will not influence responses, per the secondary reports of its announcement. Verify with OpenAI's own pages.
• Conversational formats. Platforms have talked about ads that suggest a next step within a conversation. Details are limited.

What matters for you is not the exact format, which will change, but the principles: ads are meant to be relevant to the query, labelled and distinct from the answer.

What this means for advertisers

Area	Implication	What to do
Conversion tracking	Automated systems decide where your ads appear. They learn from your conversion data	Audit tracking and goals. Import offline conversions where relevant
Landing pages	People arriving from an AI answer are often further down the decision path and expect specifics	Make pages answer the next question: price, proof, availability, process
Product data	Shopping ads depend on accurate feeds. Google says AI responses can include product listings	Fix feed titles, attributes, prices and stock. Keep them consistent with the site
Creative assets	Systems combine assets to fit a query	Provide varied, accurate headlines, descriptions and images
Match types and AI features	Eligibility is tied to broad match, AI Max and Performance Max settings	Test these with guardrails, rather than switching everything. See AI in PPC
Brand and exclusions	Automated placements can include queries you did not anticipate	Use negative keywords, brand exclusions and placement controls where offered
Measurement	Reporting by placement may be limited	Judge on blended results, incrementality tests and real sales. See attribution models
Compliance	Claims in generated or assembled ads must still be true	Review assets, and do not rely on automation to catch errors

A careful approach to testing

1. Start from solid foundations. Accurate conversions, clean feeds, strong landing pages and break-even numbers.
2. Check eligibility and settings. Read current platform help for AI placements in your country and campaign types.
3. Test, do not flip everything. Use experiments to compare setups that are eligible for AI placements with your current structure.
4. Set a test budget you can afford to lose, with caps and alerts.
5. Look at the reports you do have. Search terms, asset reports and placement data, where available.
6. Judge on business results: profit, qualified leads and revenue, plus lift tests where possible.
7. Keep notes. Placements and formats change, so record the date and the settings for each result.

What about advertising in ChatGPT?

If you are interested in ChatGPT ads, follow OpenAI's official advertising information. Points to consider:

• Availability. Reported testing began in the United States, so check whether it is available to your market.
• Landing pages. OpenAI's documentation lists OAI-AdsBot, which visits pages submitted as ads to check safety. Make sure your firewall and robots settings do not block it if you advertise. It is a separate crawler from OAI-SearchBot, which relates to ChatGPT search.
• Policies. Expect restrictions around sensitive topics and categories, and read them.
• Measurement. Plan UTM tagging and conversion tracking from the start. OpenAI says referral links from ChatGPT search include utm_source=chatgpt.com, which relates to organic referrals, so use your own parameters for paid traffic.

See our guide to ChatGPT search optimisation for the organic side.

How paid and organic visibility relate

Ads in AI answers do not replace organic visibility, and organic visibility is separate. Google's guide to its generative AI features states that SEO best practices remain relevant for organic inclusion, and ads follow ad rules and auctions. A sensible view:

• Organic earns trust and citations. Clear, accurate, specific content may be cited without payment. See our guides to Google AI Overviews and GEO.
• Paid buys presence at the moment of decision for queries where ads appear, if you meet the eligibility and win the auction.
• Together they cover more of the journey, but you should measure each separately and together.
• Zero-click effects apply to both. If answers satisfy people without a click, paid and organic traffic can fall. See our guide to zero-click searches.

Questions to ask your agency or platform representative

1. Are our current campaigns eligible for AI placements in our country? How do we know?
2. What controls do we have over where ads appear in AI answers?
3. What reporting do we get by placement, and what is missing?
4. What did you change to qualify, and what was the effect on spend and results?
5. How are we protecting brand safety and avoiding irrelevant queries?
6. How are we measuring incrementality, not just platform-reported conversions?
7. What is the test budget and the rule for stopping?

Be wary of any claim of guaranteed placements or returns.

Risks and unknowns

• Limited transparency. You may not see the exact queries or placements that triggered your ads.
• Cost. New placements can raise auction competition, or be cheap at first and then change.
• Trust. Users may react differently to ads inside conversational answers than to traditional ads.
• Policy and regulation. Rules on labelling, personalisation and sensitive topics continue to develop.
• Fast change. Formats, names and eligibility can change within months.

Common mistakes

• Switching on every AI feature without tracking or guardrails.
• Judging results from platform-reported conversions only.
• Poor product feeds and landing pages.
• Ignoring brand exclusions and negative controls.
• Treating AI placements as a replacement for organic strategy.
• Believing claims of guaranteed visibility.
• Not recording settings and dates, so you cannot learn.

A worked example: preparing a shop for AI placements

This is an invented example to show the thinking. A UK online shop sells running shoes and spends £6,000 a month on Search and Shopping ads. It wants to know whether to opt into AI placements.

1. Tracking audit. It finds that "add to basket" is counted as a primary conversion alongside purchases, which inflates conversions and misleads bidding. It changes the primary conversion to purchases with values.
2. Feed clean-up. Product titles lack gender and surface type, and many items show the wrong availability. It fixes titles and syncs stock, so shoppers and automated systems see accurate information.
3. Landing page review. Visitors arriving from specific questions, such as "best running shoes for flat feet", land on generic category pages. The shop adds short buying guides and clear size and return information to the key pages.
4. Guardrails. It adds brand exclusions, a negative keyword list and a budget cap for the test.
5. Experiment. It runs a campaign experiment comparing its current structure with a version using broader matching and AI features, splitting traffic evenly for six weeks.
6. Judgement. It compares profit, not platform conversions. The broader version brings extra sales but a higher cost per order, so the shop keeps it only for products with higher margins.

The lesson is that the preparation, tracking, feeds and pages, was worth more than the switch itself, and that the decision came from a controlled test.

A checklist before you opt in

Check	Why it matters
Primary conversions are real outcomes, tracked once	Automated systems optimise for what you count
Conversion values or offline imports are in place	Lets systems prioritise valuable customers
Product feed titles, prices and stock are accurate	Shopping and AI placements depend on feed data
Landing pages answer specific questions and load fast	People arriving from AI answers are further down the decision path
Brand exclusions and negative keywords are set	Prevents wasted spend and brand cannibalisation
A test budget and stop rule are agreed	Limits losses while you learn
You can measure incrementality or blended results	Platform reporting may be limited
Ad assets are reviewed for accuracy and policy compliance	You remain responsible for every claim

How this affects different kinds of business

Business	Likely effect	Priority
Online shops	Shopping ads and product data feed AI placements, and Google says AI responses can include product listings	Feed quality, margins and landing pages
Local services	Local intent questions are common in assistants, and Google added local business query support to some units in 2026	Business Profile, reviews, call tracking
B2B and SaaS	Comparison and "best for" questions often reach assistants first	Comparison pages, proof and lead quality signals
Publishers	Ads may fund visibility, but clicks may fall for simple questions	Original content and non-click value
Regulated sectors	Policies are stricter around health, finance and similar topics, and platforms may exclude some categories	Compliance and policy review

Creative and messaging for AI placements

When ads appear next to an answer, the person has often just read a summary. Your message works best when it adds something the summary cannot: a specific offer, a real differentiator, availability, a guarantee or a next step.

• Be specific. "Free next-day delivery on orders before 3pm" beats "great service".
• Match the question. Use headlines that reflect the problem people asked about, within the assets you supply.
• Do not restate the answer. Offer the action: compare, book, check stock.
• Keep claims substantiated. Automated systems may combine your assets in new ways, so every asset must be true on its own.
• Offer varied assets. Different angles give the system more to match.

Glossary

Term	Meaning
AI Overviews	AI-generated summaries on some Google results pages, with links and, where announced, ads
AI Mode	A conversational Google Search experience for complex questions
AI Max	Google's AI-powered option for Search campaigns, described as its primary AI campaign solution in 2026
Performance Max	A Google Ads campaign type serving across Google inventory from one campaign
OAI-AdsBot	OpenAI's crawler that checks pages submitted as ads on ChatGPT
Incrementality	The extra sales caused by advertising, compared with what would have happened anyway
Universal Commerce Protocol	An agent commerce protocol Google referred to in 2026 announcements

Planning scenarios for the next twelve months

Scenario	What happens	How to respond
Ads in AI answers expand and perform well	More queries show ads, and conversion is strong	Keep tracking, feeds and landing pages strong, and shift tested budget toward what works
They expand but perform inconsistently	Costs rise or quality varies, with limited reporting	Use experiments and holdouts. Cap spend. Prefer placements you can measure
Formats change quickly	New units replace old ones within months	Keep a testing process and notes. Do not build a strategy around one format
Regulation tightens	Rules on labelling or personalisation change	Keep claims substantiated and labelling clear. Watch announcements from regulators and platforms

Questions about measuring ads in AI answers

Question	Answer
Can I see which queries triggered my ad?	It depends on the platform and campaign type, and detail may be limited. Check current reports
How do I separate AI placement results from normal ones?	Look for placement or surface breakdowns where offered, and use experiments to compare setups
What if the platform shows great results but sales are flat?	Check overlap and attribution. Reconcile against real orders and run a holdout
How long should I test?	Long enough to collect meaningful conversions, often several weeks, and avoid changes during the test
What budget should I allocate?	A share you can afford to lose, set with caps. Increase only on evidence

Landing page guidance for visitors from AI answers

People who click an ad next to an AI answer have often already read a summary. They want the next piece of information quickly.

• Put the specific answer to their likely next question near the top: price, availability, how it works, what is included.
• Show proof: reviews, credentials and results, genuinely sourced.
• Make the next step obvious and quick on mobile.
• Keep the page consistent with the ad, since automated systems may assemble the ad from your assets.
• Avoid popups that interrupt the first view.
• Make sure OAI-AdsBot and other ad review crawlers can access the page if you advertise on platforms that use them.

Why the details keep changing, and how to keep up

If you read three articles about ads in AI answers, written a few months apart, they may describe different formats, different eligibility and different countries. That is not because the writers are careless. It is because the products are changing quickly, and the companies announce features in stages: first as announcements at events, then as tests with a subset of advertisers, then as wider rollouts, then, often, with changes after feedback. Terms also shift. A feature announced under one name may later be folded into another campaign type.

For anyone responsible for budget, the practical response is to build a habit of checking primary sources and keeping your own record. Here is a simple routine.

Each month, open the official release notes or announcements pages for the platforms you use: Google Ads help announcements, and Google's Search Central and ads blogs; OpenAI's help centre and crawler documentation; the platform pages for any other AI product where you advertise. Read what changed, note the date and note whether the change applies to your country and campaign types. Write a short line in a shared document: "On this date, platform X announced Y. Applies to us: yes or no. Action: test, wait or ignore."

When you start a test, record the date, the campaign settings, the budget, the hypothesis and the tracking set-up. When the test ends, record the result and your interpretation. Six months later, when a colleague asks why you opted in or out, you will have an answer. This is dull work, but it is the difference between a strategy and a series of reactions to announcements.

Also be sceptical of commentary. Some sources, including some articles that cite statistics about ChatGPT ads or AI Overview ad performance, rely on single studies, small samples or unverified screenshots. Others present early test results as if they were established patterns. Where a claim matters to a decision, trace it to a primary source, such as the company's own announcement or documentation, and note the date. Where you cannot trace it, treat it as a hypothesis to test, not a fact.

Finally, remember that your own data is the best evidence. A small, well-run experiment in your account, with caps on spend and a clear stop rule, will teach you more about whether ads in AI answers work for your business than any article, including this one. The goal is not to be first. It is to be informed, to avoid expensive mistakes and to adopt what works for you when the evidence is there.

A closing reminder

Whichever platform you test, write down your hypothesis, budget and stop rule before you start, and check results against your own sales. That habit protects your budget while the market settles.

Where we can help

We are a digital marketing agency in Manchester, UK and Mumbai, India. Our AI PPC services and Google Ads management include tracking audits, feed checks and controlled tests of new placements. Contact us to discuss a careful approach.