On this page11
- What Google Announced
- How the Report Counts
- What the Report Does Not Show
- Why Impressions and Answers Can Disagree
- What Brand Mention Monitoring Software Adds
- Common Mistakes When Reporting AI Impressions
- A Monthly Client Report That Uses Both
- Use Both Reports Together
- A Note on the Opt-Out Control
- What to Tell Clients
- Frequently Asked Questions
Google's Search Console now has a Generative AI performance report. It counts how often links to your site appear in AI Overviews and AI Mode, by page, country, device and date. It does not read the answer itself. Brand mention monitoring software reads what AI says about a client. Agencies need both to report with confidence and evidence.
What Google Announced
Google announced the new reports on June 3, 2026. Mrinalini Loew, General Manager of Google Search Ecosystem, wrote the announcement. Google updated the post on August 31, 2026. It now says the features have reached all websites worldwide. Search Engine Land reported the global rollout the same day.
The details below come from Google's announcement and its own help page for the report. Nothing here is a guess about how the report works.
| Item | What Google says |
|---|---|
| Name | Generative AI performance report (Search) |
| Features covered | AI Overviews and AI Mode. Google may expand the list. |
| Metric | Impressions: "how many times links to your site were shown to a user in a generative AI feature" |
| Dimensions | Pages, countries, dates and devices |
| Search types | Web text, and Web multimodal such as image-based searches |
| Availability | All websites worldwide as of August 31, 2026 |
| Related control | A toggle that lets owners choose whether their site appears in, and helps ground, these AI features |
This is good news for agencies. Google itself now reports how often a client's pages reach an AI answer. It is a first-party number, and every client with Search Console can open it.
How the Report Counts
The counting rules matter when you present the numbers. Google's help page lists several.
| Rule in the help page | What it means for a client report |
|---|---|
| Chart totals count by property | Two results from one site in the same feature count as one impression in the chart |
| Page data counts by page | Chart totals and table totals can differ |
| Only the Web search type | The report does not cover every Search Console search type |
| Search Labs is excluded | Experimental features do not appear |
| Discover has its own report | Do not add the two together |
| 1,000-row limit | Large sites see a cut-off in the table |
| Newest data may be preliminary | Recent days can change after you report them |
| Report may not appear | A site with too few AI impressions may see no report |
Read the first two rows twice. A client with ten pages shown in one AI Overview can still see a small chart number. State that rule in the report notes before the client asks.
What the Report Does Not Show
Google's help page lists impressions as the metric. It lists no click figure. Search Engine Land makes the same point about clicks. The help page also does not list the answer text, the question a user asked, or any assistant outside Google.
That is not a flaw. A report built from Search Console data counts what happens to a site. It was not designed to read what an assistant says. Those are two jobs. Brand mention monitoring software does the second one.
| Question a client asks | Search Console AI report | Reading the answers |
|---|---|---|
| Was my page shown in an AI answer? | Yes, counted as an impression | Yes, if the answer shows it |
| What did the answer say about us? | Not listed in the report | Yes, the stored answer text |
| Was the claim correct? | Not listed in the report | Yes, checked against the client's facts |
| Who else was named beside us? | Not listed in the report | Yes, rivals in the same answer |
| What do ChatGPT and Perplexity say? | Outside its scope | Yes, across nine assistants |
| Did visits follow? | No click metric in the report | Connect GA4 for visits |
Why Impressions and Answers Can Disagree
An impression says a link was shown. It does not say the answer around it was right. The next case is illustrative. It is not measured data.
Picture a client that changed its price last spring. Its pages show the new price. The Search Console report shows steady AI impressions for those pages. The report looks healthy. A buyer asks an assistant about the client and reads the old price from an old third-party page. The impression was real. The answer was wrong.
Only a stored reading of the answer would show that gap. This is why AI visibility tracking needs the answer text, not only the count. SKYA explains the wider problem in its guide to how AI platforms cite sources.
The gap works in the other direction too. A client can have few AI impressions and still be named correctly in answers that cite a third-party page. Impressions measure links to the client's own site. Answers draw on many sources. Read both views before you tell a client the work is succeeding or failing.
What Brand Mention Monitoring Software Adds
A good tool fills the gaps in the table above. Five requirements separate a real reading from a guess.
| Requirement | Why it matters | How SKYA handles it |
|---|---|---|
| Reads many assistants daily | Buyers use more than Google | Nine assistants every day: ChatGPT, Gemini, Perplexity, Claude, Google AI Overview, Google AI Mode, DeepSeek, Grok and Copilot |
| Stores the answer text | Proof beats memory | A stored reading behind every number, with an evidence panel |
| Checks accuracy | A wrong mention can cost more than none | A hallucination audit on every paid plan |
| Admits what it cannot read | An invented number misleads the client | No answers read means no figure. A run that measured nothing gives its credit back |
| Reports by client | Agencies must show change | Multi-client workspaces and white-label PDF reports |
SKYA also connects GA4 and Search Console. That puts Google's own numbers beside the stored answers in one view. The agency then reports the count and the content together. SKYA publishes how every number is calculated.
Common Mistakes When Reporting AI Impressions
A new report invites quick conclusions. Five habits cause most of the trouble.
| Mistake | Result | Better move |
|---|---|---|
| Calling impressions visits | The client expects traffic that never arrives | Say "shown", not "visited". Connect GA4 for visits |
| Adding Discover to Search | Double counting | Report them separately, as Google does |
| Treating the newest days as final | Numbers shift after the meeting | Wait for preliminary data to settle |
| Reading one page alone | The pattern is missed | Read the top and bottom pages together |
| Skipping the answer text | Wrong claims go unseen | Attach the stored answer to each claim |
None of these mistakes comes from the report. They come from using one number to answer a question it was not built to answer.
A Monthly Client Report That Uses Both
A client wants one page that says what happened and what comes next. The layout below keeps each number tied to its source.
| Section | Source | What the client learns |
|---|---|---|
| AI impressions trend | Search Console AI report | How often pages reached AI Overviews and AI Mode |
| Mentions and citations | Daily answer readings | Whether the client is named and cited |
| Accuracy | Hallucination audit | Which claims were wrong and which were fixed |
| Rivals in the same answers | Stored answers | Who appears beside the client |
| Visits | GA4 | What happened after the answer |
| Work done | Agency notes | Which pages and sources changed |
The first column answers "how often". The rest answer "what was said" and "what did it do". A report with only the first column leaves the client guessing.
Use Both Reports Together
The Search Console report and a daily reading do not compete. Each fills in what the other cannot. A simple routine joins them.
-
Open the Generative AI performance report monthly. Note the top pages and the countries.
-
Pick ten buyer questions for each client. Include questions that map to those top pages.
-
Read the answers daily across the assistants. Store each answer.
-
Compare the two. A page with many impressions and a wrong answer goes to the top of the fix list.
-
Fix the sources. Correct the client's pages, align third-party mentions and add structured data. Schema markup for AEO explains which types fit which pages.
-
Report both numbers. Show the impression count and attach the stored answer.
This routine gives a client a count Google stands behind and an answer the client can read. Together they make AI visibility tracking credible. Either one alone leaves a hole.
A Note on the Opt-Out Control
The announcement also introduced a control. Google says sites that opt out "will not receive traffic or impressions from our generative AI features." Google also says the control will not be used as a ranking signal for results outside those AI features.
| Question | Plain answer |
|---|---|
| What does the control do? | It decides whether the site appears in, and helps ground, AI Overviews and AI Mode |
| What happens if a client opts out? | No traffic or impressions from those AI features, per Google |
| Does it affect ordinary rankings? | Google says it is not a ranking signal outside the AI features |
| Should an agency change it? | SKYA's view: only with the client's written approval, and only after a baseline reading |
The last row is SKYA's advice, not Google's. Opting out removes the client from those answers. A baseline reading first shows what the client would lose.
What to Tell Clients
| Say | Avoid |
|---|---|
| Google now reports how often your pages appear in AI Overviews and AI Mode. We read what the answers say. | Search Console tells us everything AI says about you. |
| The report counts impressions. It does not show clicks. | These impressions are visits. |
| We store the answer text so you can read it yourself. | Trust the dashboard. |
| We watch the trend over weeks, not a single day. | Impressions will rise every month. |
The pattern is simple. Use Google's number for what it measures. Use stored answers for what it does not. A client who sees both trusts the report more.
Frequently Asked Questions
What is the Generative AI performance report in Search Console?
It is a Search Console report that shows how often links to your site were shown in generative AI features on Google Search. Google says it covers AI Overviews and AI Mode, with impressions by page, country, date and device.
Does the Search Console AI report show clicks?
No. Google's help page lists impressions as the metric, and Search Engine Land notes the report has no click data. Connect GA4 to see visits.
Can Search Console show what AI says about my client?
Not from what Google lists. The report counts impressions of links. To read the answer text and check it for accuracy, an agency needs stored readings of the assistants.
What is brand mention monitoring software?
Brand mention monitoring software reads what AI assistants say about a client and records whether it is named, cited and accurate. SKYA does this daily across nine assistants and stores the answer behind every number.
How do I start AI visibility tracking for a client?
Write ten buyer questions, read the answers across the assistants daily and store each one. SKYA's free AI Visibility Scan gives a first report in about two minutes with no sign-up.
Make AI Mention Your Brand. Start a 7-day free trial. No card required.
Written by
SKYA Editorial Team
The SKYA Editorial Team publishes field-tested research on Answer Engine Optimization, Generative Engine Optimization and AI Visibility. Every article is reviewed by senior practitioners from the SKYA product and research groups before publication.
Related posts
Keep exploring the SKYA Journal.
