There is no "show my ads in AI Overviews" checkbox anywhere in Google Ads, and that absence is the most important fact about the format. Eligibility is automatic: if you run Search, Shopping or Performance Max, your ads can already serve inside Google's AI answers. Which means showing up well is not a media-buying decision — it is an account-quality outcome. That is good news, because account quality is something you can actually work on.
Where ads appear in AI Overviews and AI Mode today
In AI Overviews, ads serve above, below or within the Overview itself, on mobile and desktop, currently in English across twelve countries including the US, Canada, Australia, India and Singapore. Google's help documentation states the serving logic plainly: both the user's query and the content of the AI Overview are considered. That second input is new. Your ad can match the answer Google composed, not just the question the user typed.
AI Mode — the fully conversational search surface — is running ads in US testing, matched on conversational context that accumulates across a session rather than on a single query. Third-party tracking through 2026 puts ads in roughly a quarter of AI Overview results, several times more than a year earlier, so this is no longer an edge placement. The direction of travel is unambiguous: more AI surface, more ad slots inside it.
Which campaign types qualify automatically
| Campaign type | AI Overviews | AI Mode (testing) | Your control surface |
|---|---|---|---|
| Search (text ads) | Eligible automatically | Eligible in tests | Keywords, negatives, assets |
| Shopping | Eligible automatically | Eligible in tests | Feed quality, titles, price data |
| Performance Max | Eligible automatically | Eligible in tests | Asset groups, feeds, exclusions |
| Display / Video | Not the inventory source | — | — |
Two implications hide in that table. First, Performance Max quietly becomes an AI-surface play: its asset groups and feeds are exactly the raw material the new placements consume. Second, your fences still work — negative keywords, brand exclusions and feed hygiene police AI-surface matching just as they police classic Search, and they are currently the only policing you get.
Why asset and feed quality decide visibility
A classic search results page ranks ads; an AI answer composes with them. The system is choosing an ad to sit inside or beside a synthesised paragraph, so it favours material that fits: headlines that state what the product is in plain words, descriptions that answer rather than tease, Shopping titles with concrete attributes, prices and availability that are current, landing pages that deliver what the ad names.
The practical checklist writes itself, and a connected AI agent can run it against your real account:
- RSA coverage — every active ad group has a strong ad with full headline coverage and no dangling "learn more" vagueness. Drafting against real queries is covered in writing RSAs with AI.
- Feed truth — Shopping titles, prices, availability and GTINs current and specific; a feed error is now an answer error.
- Landing page match — the page says what the ad said, quickly; AI surfaces sit closer to the consideration moment, and a mismatch wastes a more expensive click.
- Query hygiene — conversational, longer queries are exactly where loose matching drifts; the negative list earns its keep here.
None of this is new advice. What is new is the penalty for ignoring it: on a ten-blue-links page a mediocre ad underperforms, while inside an answer it may simply never be chosen.
How AI surfaces change the queries you buy
The second-order effect matters as much as the placements. People type differently into an answer engine than into a search box: longer queries, full questions, stacked constraints — "waterproof hiking boots wide fit under a week's delivery" instead of "hiking boots". As AI Overviews and AI Mode absorb more of the journey, that phrasing shows up in your search terms report, and it interacts with every matching decision you have made.
Three consequences to manage deliberately. Exact-heavy accounts will feel reach erode first, because enumerated keywords cannot cover conversational phrasings — this is the demand-side argument behind broad match and AI Max adoption. Negative lists built against short queries need re-reading, because a conversational query can drift through a fence built for two-word tokens. And intent gets more legible, not less: a stacked-constraint query tells you precisely what the buyer needs, which makes query mining more valuable per row than it has ever been. Watch the drift monthly; it is the clearest signal of how much of your market has moved surfaces.
What reporting exists (and what does not)
As of mid-2026, Google does not break out AI Overview or AI Mode placements as their own reporting row — they fold into standard Search metrics. You cannot yet ask "what did AI surfaces cost me and what did they convert". Honest measurement today is inferential: watch query-length and query-shape drift in the search terms report, watch top-of-page metrics on affected campaigns, and annotate your change history when Google expands eligibility so later analysis has its landmarks. When placement segmentation does arrive, accounts that annotated will be able to read it; accounts that did not will be guessing at their own history.
What it costs, honestly
There is no separate price sheet for AI-surface placements — the same auctions, budgets and bid strategies govern them. Early third-party coverage describes stronger engagement at premium click costs, which is believable for a placement closer to a composed answer, but treat any specific figure with suspicion: the format is expanding monthly, auction density is still settling, and your query mix decides your outcome. The defensible position is procedural, not numerical — measure your own affected campaigns before and after eligibility expansions, rather than importing an average from someone else's account.
The procedure is worth spelling out, because it is cheap and almost nobody runs it. Keep a dated log of AI-surface milestones as Google announces them — eligibility expansions, new countries, reporting changes. When each lands, have a connected agent snapshot the affected campaigns for the trailing four weeks: cost per click, top-of-page share, conversion rate, query-length distribution. Repeat four weeks after. The deltas will not be laboratory-clean, but a consistent pattern across campaigns and milestones is evidence — your evidence — and it will be ready months before official reporting catches up.
Preparing an account without chasing a checkbox
The strategy reduces to one sentence: make the account so legible that a machine can quote it mid-answer without embarrassing anyone. Run the four-point hygiene list quarterly. Keep feeds truthful weekly. Let your negative keywords absorb the conversational drift. And track the year's surface changes as they land — the running inventory is in what actually changed in Google Ads in 2026.
The accounts that win AI-surface placement will not be the ones that found a secret setting. There isn't one. They will be the ones whose ordinary hygiene was already answer-shaped when the answers arrived — and whose operators could prove it, because they measured their own campaigns while everyone else waited for a reporting row.