If your Google Ads account runs on exact match keywords and manual bidding, it cannot show ads in AI Search. Ads Liaison Ginny Marvin confirmed the eligibility requirements have not changed: getting into that inventory takes AI-powered targeting. Broad match qualifies. So does keywordless targeting through AI Max, Performance Max, Shopping, Dynamic Search Ads, or Smart Bidding.
Marvin set this out in an Ads Decoded newsletter and video following Google Marketing Live. Search Engine Journal covered the clarification on 13 July.
The eligibility rule, plainly
No new rule was announced. That is the news. Advertisers kept asking whether AI Search had opened up as separate inventory with separate requirements, and the answer is that the existing requirement still stands.
Which draws a line through a very common account setup. Tight exact-match ad groups on manual CPC, the structure plenty of advertisers still trust precisely because it is controllable and easy to audit, has no route into AI Search placements at all. Control is the thing being traded.
That is a real decision rather than a technicality. It is worth being clear-eyed that Google has an interest in the answer, since every qualifying option on that list hands more of the targeting to Google. The rule is still the rule.
The relevance bar is higher there
Marvin also said the relevance bar in AI Search sits higher than in standard search. Ads are matched against Google's understanding of intent, drawn from both the query and the content of the response.
That second half deserves a re-read. The response is an input to the match. The same query can produce different responses, so the ad is being matched to the whole exchange rather than to a string of text a user typed. Keyword-level control has been loosening for years; this describes a system where the keyword is not the unit of matching at all.
The practical consequence is that the answer text on the page becomes part of what an advertiser is bidding into. That puts paid search and generative engine optimisation closer together than most account structures currently assume.
Qualified Future Conversions is a prediction
Marvin also detailed Qualified Future Conversions. It estimates conversions Google expects to occur up to 180 days after an ad interaction, built by combining early signals such as branded searches with historical data.
Estimated. Predicted. Not measured, not observed, not banked. QFC is a model output about a future that has not happened yet, and it will be sitting in reports next to columns that count things that did happen. That is the whole risk with the metric, and it has nothing to do with whether the model is any good.
The attribution gap, in Google's own numbers
Google says roughly 70% of conversions on standard campaigns land inside standard attribution windows. On Performance Max that falls to 50%. On Demand Gen, 40%.
Turn those around and they read differently. On Demand Gen, about 60% of the conversions Google believes are happening fall outside the window being reported on. Half of them on Performance Max. Around 30% on standard campaigns.
This is useful, because it is an argument agencies have been making for years without a citation. When a Demand Gen campaign looks flat next to brand search, some of that gap is measurement timing rather than performance. The figure now has Google's name on it.
The caveat carries equal weight. Those percentages come from Google, describe Google's own products, and happen to flatter exactly the products that look worst on reported conversions. That makes them a reasonable directional input and not an independent audit. Quote them as Google's numbers, because that is what they are.
What this means for Thai marketers
Thailand was not mentioned in any of this. The eligibility rule is a product rule, so it applies to accounts here the same way it applies anywhere.
- If you have been holding the line on exact match and manual CPC, that position now has a named cost: no AI Search inventory. Whether the cost matters depends on how much of your volume you expect to end up there.
- Broad match plus Smart Bidding is the entry ticket. It is also the setup that fails loudest when conversion tracking is thin, which is an argument for fixing tracking first rather than for skipping the move.
- Do not report QFC to a client as conversions. It is a forecast. Give it its own line, label it as predicted, and keep it out of the cost-per-conversion maths.
- The 70/50/40 split belongs in the deck for the meeting where someone asks why Demand Gen looks weak. Attribute it to Google.
- A 180-day prediction window is longer than most reporting cycles and most retainers. Decide now how you will handle a number that cannot be checked until well after the campaign is judged.
For accounts that have been run conservatively, the honest read is that the conservative structure now has a ceiling that did not exist before. Moving off it is a Google Ads management decision about tracking quality and risk tolerance, not a philosophical one about automation.
FAQ
Can exact match keywords show ads in AI Search?
No. Marvin said eligibility requires AI-powered targeting: broad match, or keywordless targeting through AI Max, Performance Max, Shopping, Dynamic Search Ads, or Smart Bidding.
Did the AI Search ad rules change?
No. Marvin said the requirements have not changed.
Is Qualified Future Conversions a count of real conversions?
No. It is a predictive estimate of conversions up to 180 days after an ad interaction, based on early signals such as branded searches combined with historical data.
How are ads matched in AI Search?
Against Google's understanding of intent, based on both the query and the content of the response. Marvin said the relevance bar is higher than in standard search.
What share of conversions falls outside standard attribution windows?
By Google's figures, around 30% on standard campaigns, 50% on Performance Max and 60% on Demand Gen.
Where this leaves the account
Two things landed at once here, and they point the same way. Access to new inventory requires giving up keyword control, and the metric offered in exchange is a prediction rather than a count. Whether that trade is worth making is an account-level question about how much you trust your own conversion data. If that question is open in your account, it is worth answering before the next budget cycle rather than during it.
