Google AI Mode Ads: What Runs, Who's Eligible, How to Check

You do not need a separate campaign. AI Mode is Google Search's conversational mode, where a question gets a written answer and follow-up questions. Ads reach it through campaigns you already run. Google's guidance for the new formats is to build on AI Max for Search, AI Max for Shopping and Performance Max, and ads carry a Sponsored label.

Three terms matter here:

• Performance Max (PMax) is Google's automated campaign type that runs across its channels.
• AI Max for Search is a setting on Search campaigns that widens how queries are matched to your ads.
• Broad match is the loosest keyword setting. It lets Google match related queries, not only your exact words.

The formats Google announced on May 20, 2026, with their status as of October 2026:

Google's Ads Liaison Ginny Marvin ties the newest formats to the automated products: "AI Max and PMax for Search are designed to capture complex conversational intent and access our latest ad formats (like Highlighted Answers)."

The practical consequence is that you cannot switch AI Mode on. You can only be eligible, so the first job is knowing whether you are.

Performance Max, Shopping, and Search campaigns using broad match qualify, and Google's newer formats are built on AI Max for Search. As of September 1, 2026, Google automatically upgrades Search campaigns that use campaign-level broad match or automatically created assets to AI Max for Search. A Search campaign you never touched may now be an AI Max campaign.

Google announced the change in August 2026: "Starting September 1, 2026, campaigns using automatically created assets (ACA) and/or campaign-level broad match setting will automatically be upgraded to AI Max for Search campaigns." The original eligibility of Performance Max, Shopping, and broad-match Search campaigns came with the May 2025 announcement, which described ads as part of a test. Our recommendation is to open your Search campaigns and check which ones carry AI Max settings before you read any AI Mode result.

Exact and phrase match keywords joined on September 4, 2026. Marvin called it "a small experiment that recently kicked off", and said exact and phrase match keywords can serve text ads in AI Mode in cases of "explicit and direct user intent." Before this experiment, exact and phrase match campaigns could not show in AI Mode. Anthony Higman, the advertiser who spotted the change, argued it is bigger than Google's framing. Because Google calls it a small experiment, do not build a budget on it. Our recommendation: keep your exact and phrase match structure as it is and test with a small budget.

For control, Google's help page for AI Overviews ads states that you cannot target the placement, cannot opt out, and get no segmented reporting. Assume AI Mode works the same way and measure it yourself.

No source shows that buying an AI Mode ad puts your brand in the answer. Google labels the ad as a separate Sponsored unit, so the answer's text and citations come from a different process. Our recommendation is to measure both rather than assume one lifts the other.

The labeling is not always obvious to a reader. In one early test reported by MediaPost, a Sponsored listing appeared in the middle of a list of coffeemaker recommendations. When a sponsored entry sits in the same list as unpaid recommendations, your unpaid presence matters twice: it earns trust, and it is what a reader compares your ad against.

In our research at Citation Research (Study 007, fieldwork October 5, 2026), we asked 150 questions from India across 10 topics. Google AI Mode cited Reddit, YouTube or Quora in 117 of the 150 answers, and ChatGPT did so in 3. We have not run the US version of that question set, so treat this as a signal to check community sources, not a US rate.

Our inference, beyond what the study measures, is that an AI Mode answer can lean on threads and videos you cannot bid on. A brand that only buys the ad could win the slot and still be missing from the part of the answer that carries the recommendation. The sheet below tests whether that is true for you.

Run the same prompts several times and record, for each one, whether an ad appeared, whose it was, and whether you were cited. It takes a spreadsheet and a fixed setup. Start from the searches you already pay for, then widen to the questions buyers ask before they search for you.

The sheet. One row per prompt:

Run rules. Run each prompt 3 times and count a cell as Y only if it appears in at least 2 of the 3 runs. That 2-of-3 rule is our recommendation, not a tested standard, and it exists because AI answers vary between runs. Repeats add reliability, not sample size.

How many prompts. Our model uses the standard proportion formula: n = z² × p × (1 − p) ÷ e². With 95% confidence (z = 1.96) and the cautious assumption p = 0.5:

• For ±10 percentage points: 1.96² × 0.25 ÷ 0.10² = 0.9604 ÷ 0.01 = 96.04, so 97 prompts.
• For ±5 percentage points: 0.9604 ÷ 0.0025 = 384.16, so 385 prompts.

The margin applies to each rate separately and is widest near 50%. It assumes independent prompts in one setup, and your rates describe your prompt set, not all of Google. The workload is real: 97 prompts at 3 runs each is 291 runs, and 385 prompts is 1,155. As a first pass, our recommendation is 20 to 30 prompts to learn the sheet; that is too small to put a rate on.

The rates:

• Ad rate = prompts with any ad ÷ total prompts.
• Your-ad rate = prompts with your ad ÷ total.
• Cited rate = prompts where you are cited ÷ total.

What to do with them. In this table, high means 50% or more of prompts and low means under 50%. Those cut-offs are our recommendation.

To connect the answer to revenue, tag paid clicks with UTM parameters, the tags added to a link that label its source, and read AI referrals in analytics. Our guide to tracking AI traffic in GA4 covers the channel group and the regular expression (regex) test. Choosing which prompts go on the sheet is covered in prompt research.

We track the answer side. PageLens.ai publishes this page and sells AI answer tracking. On our Launch plan we run your prompts through ChatGPT, Google AI Mode and Perplexity, and record which brands each answer names and which pages it cites. The ad columns on the sheet are a manual check; the named and cited columns are what we track.

As of October 2026, Launch costs $299 a month and tracks 100 prompts weekly (300 AI answers a week), so a 97-prompt sheet fits and three weekly runs take three weeks. Growth costs $699 a month and tracks the same 100 prompts daily (500 AI answers a day), so three runs take three days. A 385-prompt sheet is larger than any standard plan (Enterprise tracks 200), so it needs a custom plan or prompts rotated across weeks. Growth also adds Gemini and Grok, which are separate engines from Google AI Mode. Annual billing saves 8% (pricing). You can also see how AI engines describe your brand in the free AI visibility audit.

A single run tells you little, so we repeat prompts and compare over time; the method is in tracking AI citations over time. Google AI Overviews, the summary shown above regular results, carry ads too, but PageLens.ai does not track Overviews as a separate engine. For getting cited there, see how to rank in Google AI Overviews.

To set up your prompts and run the sheet on a tracked basis, book a demo.