Myths About AI Impressions: What Google's New Report Does Not Prove

Answer first: Google's generative AI report proves that your pages appeared inside AI Overviews and AI Mode, and very little beyond that. It does not prove clicks, rank, or revenue, so the safest way to use it is to know its limits before you draw any conclusion.

Every site owner now has access. Google launched the dedicated reports on June 3, and a note in its documentation says every website worldwide had access by August 31, 2026. Below are the myths that tend to spread once owners open the new view, followed by what the data really supports.

Myth one: AI impressions are extra traffic on top of your totals

They are not extra volume. Google's documentation says the generative AI report draws on the Performance report's Web search type, which makes it an isolated slice of numbers you already had.

That detail is useful once you see it. Divide your AI impressions by your total web impressions for the same window, and you get a rough share of visibility inside AI answers.

Myth two: an impression means someone read your citation

An impression only means the link reached the screen. Standard Search Console rules apply, so a link has to be scrolled into view or expanded before it counts.

John Mueller confirmed this in June 2026. When a user must activate a link to reveal it, the impression counts only from that moment forward, which matters a great deal inside collapsed AI answers.

Myth three: a low count means Google ignored the page

A small number can have quiet causes. Your citation may have sat behind a click that nobody made, so the page was used but never displayed.

Some pages also aim at searches where Google never shows an AI answer at all. Measure each page by the purpose it was written for, not by an AI count it could never earn.

Myth four: the report shows clicks, queries, and position

It shows none of the three. The reports break out impressions from AI Overviews, AI Mode, and the generative AI features in Discover by page, country, and date, and the Search version adds devices.

At launch Google left out click-through rate and query data, and it said more metrics would arrive over time without naming a date. Average position is missing for a sensible reason, because a generative answer has no fixed first or third spot.

Citations can show up as inline links, side panel cards, or footnotes that expand. A single position number would describe none of those layouts honestly.

Myth five: an odd September chart means you lost ground

A strange shape early in September may not reflect your site at all. Industry coverage that month noted that the new reports kept growing, and that a bug had hit the report a couple of weeks earlier.

Before you rewrite pages over one bumpy stretch, compare a wider date range. A reporting glitch and a real decline can look identical in a single screenshot.

Myth six: opting out keeps you safe

The opt-out mostly takes visibility away. The same release added a Search Console toggle that blocks your content from AI search features, and Google says opted-out sites get no traffic or impressions from those features.

Google adds that the control plays no role as a ranking signal in regular web search. So it neither protects your normal rankings nor helps them, and for most small businesses it simply removes a chance to be cited.

Finer choices may come later. Under a CMA timeline, Google has until March 2027 to offer controls at the page level instead of the current property setting. The regulator also expects click-through data, and Google must file compliance reports every six months in the first year.

Myth seven: impressions do not matter if sessions stay flat

Sessions that hold still are exactly why impression share deserves a place in your tracking. One analysis of AI Mode found that roughly 4.5 percent of sessions finish with a click to an outside website, compared with about 24 percent for traditional search.

The AI surfaces are also taking up more of the page. For some queries, Google confirmed, AI Overviews can grow into a full detailed response with no "Show more" button, which pushes core results lower.

September coverage also reported that Google is moving AI Mode into AI Overviews within Search, with AI Mode partly powered by Gemini 3.7 Flash. When fewer people click, being named inside the answer carries more weight.

Myth eight: every page should earn AI impressions

Different page types behave differently, and that is normal. Blog archives usually behave unlike service pages, pricing pages, or clear how-to answers, so compare like with like.

A page with zero AI impressions is not automatically a failure. It may serve a search that simply does not produce an AI answer.

Myth nine: a spike from another country signals local demand

Country data can mislead a local business. If you serve one metro area, a jump in AI impressions from another country says little about the customers who can actually hire you.

Filter by country and by device before you celebrate or worry. The narrower view tells you far more about the buyers you serve.

What the data can still do for you

Incomplete data still beats guessing, as long as you read it for what it is. A short routine turns the limits into a working baseline:

• Check both the Search report and the Discover report under the Generative AI view.
• Export three months of AI impressions by page, and save the file as your starting point.
• Pair impressions with referral sources in your analytics, for example AI assistants and Google's own AI surfaces, to set real sessions beside the counts.
• Read a full month of data before you even think about the opt-out toggle.

Pages that collect AI impressions tend to share a few traits worth copying. They lead with a direct answer inside the opening two sentences, use short sections under plain headings, and state checkable facts such as licensing, warranty terms, hours, staff names, service areas, and pricing structure.

They also show a visible last updated date with current figures, and they list the same business name, address, and phone everywhere. Clean technical basics let crawlers read them quickly.

The bottom line on limits

This is the first free way for small businesses to measure visibility inside AI answers, and it is still incomplete. Treat it as a visibility signal, keep your baseline, and let real changes on your pages prove themselves over time.

Want help separating the signal from the myths in your own numbers? Contact AppWT with your Search Console access, or set up a call, and AppWT will review the report with you and map it to your service pages.