Track your visibility in Google Gemini grounding

本文目前只有英文版。站点其余部分已汉化，这篇内容尚未翻译。

Google has more than one AI answer surface, and the difference between them is not a detail. This page is about one of them: Gemini answering a question with Google Search grounding turned on. It is not about the AI Overview block on the search results page, which TriRank tracks separately and never mixes with this one.

If you only remember one thing from this page, make it that separation. It is the reason our column is labelled "Google (Gemini grounding)" rather than something shorter and more marketable.

Why the two are kept apart

When a chart's column keeps its name and quietly changes its source, every historical point in it becomes a lie — the line looks continuous and the thing it measures has moved. This engine and Google AI Overviews are stored under separate keys, forever, precisely so that cannot happen. No history was migrated between them, nothing was backfilled, and neither one was folded into the other when the second arrived.

The practical version: if you see both columns and they disagree, that is not an inconsistency to be reconciled. They are two different questions with two different answers, and both are true.

How the probe works

Each tracked query goes to Gemini with Google Search grounding enabled. The response comes back with an answer plus a set of grounding sources — the pages the model actually retrieved to build it.

There is a wrinkle here that no other engine has. Gemini does not hand back the destination URLs directly; it hands back redirect URLs on Google's own domain. A checker that matched those against your site would find zero citations for every site, forever, and the failure would look exactly like an honest run of zeroes. So each source is followed through its redirect to the page it really points at, up to a bounded number of sources per query and with a tight timeout on each hop, and the matching happens on the real destination.

The full grounding evidence is stored alongside the verdict, so a cited or not-cited call can be re-examined afterwards rather than taken on trust.

What "cited" means here

A query counts as cited when one of the resolved destination hosts is your domain exactly or a subdomain of it, after lowercasing and dropping a leading www..

The answer prose is never read for brand mentions. If Gemini describes your product in a sentence with no retrieved page behind it, that is the model's training data talking, and it is not scored as a citation.

What this cannot tell you

It is not what a searcher sees on google.com. That surface is the AI Overview, and it is tracked in its own column. This one tells you how you fare when Google's model researches a question conversationally. Those two things correlate loosely at best.

Grounding is not guaranteed to be broad. The model decides how much to retrieve. A query where it grounded lightly and a query where it grounded heavily and skipped you are different situations that produce the same not-cited cell. The stored source list is where you can tell them apart.

One reading is noise. Live retrieval varies between runs. The weekly series is the measurement; a single probe is not.

Model versions move underneath you. The grounded answer you get today comes from a specific model version, and Google ships new ones on its own schedule. Comparing a reading from this quarter against one from two quarters ago is comparing two slightly different machines.

There is no position number here, and there is not one to have. A results page has ten slots, so a rank is a row you either occupy or you do not. A synthesized answer is prose that names a few brands and links a few pages, and there is rarely a numbered position to track inside it. What this check reports is therefore presence and citation, not a row number, and nothing on this page converts into one. Where you would want a rank, the measurement doing that job is the trend — how often this engine cited you across weeks of the same questions, rather than where you placed inside any single answer. The word "rank" is still settling in this market: we use it in the sense set out under AI search ranking, and AI search rank tracking covers what tracking it involves.

There is no published cross-brand benchmark for this engine. What runs on your account is your own sweep: your tracked queries, through this engine, on the same weekly cadence as the other four. What does not exist is a public study across many brands that would let you ask how this engine behaves in general — how often it cites anyone, or which kinds of sites it tends to reach for. No number on this page comes from one, and we are not putting a date on when there will be one. The distinction is worth stating plainly, because "we do not measure this engine" and "we have not published a study about this engine" are different claims and only the second is true.

What to do with the result

Look at the resolved source list for the queries where you are absent. Because these are real destination URLs rather than redirects, that list is directly useful: it is the set of pages Google's model chose to build an answer from.

The pattern worth checking first is structural. Grounded answers pull heavily from pages that state a claim plainly and support it — the shape that generative engine optimization describes. Pages that bury the answer under preamble tend to be retrieved and then not used, which looks identical to not being retrieved at all from the outside.

The free audit runs this probe against your domain and shows the sources it resolved, including the Search Suggestions element Google requires be displayed with a grounded result. For the definition side of this engine, see Gemini Search.