How to know if you appear in Google's AI Overviews

Search Console does not separate AI Overviews: why average position cannot tell you if you are cited

The first instinct of anyone wondering how to know if they appear in AI Overviews is to open Google Search Console and look for the report. There is none. Google confirmed that impressions and clicks generated inside the AI summary are counted in the performance report, but it never shipped a search type, a tab or a filter that isolates them from the plain organic result. They are in there, blended, unlabeled.

That has an uncomfortable consequence for anyone measuring seriously: the average position of a query may be averaging two surfaces that have nothing in common. An impression of your blue link at position 8 and an impression of your domain cited inside the AI block, which visually sits above everything, go into the same pot. When the average drops you cannot tell whether you lost organic position, gained summary presence, or both at once in opposite directions. The arithmetic behind that metric, and why it rarely matches what you see when you search, is covered separately in the breakdown of Search Console average position versus real position.

There is one indirect signal you can read, and it is worth knowing how to interpret it without overreading it. If impressions on a query hold steady or rise while clicks fall and your position has not moved, it is reasonable to suspect that a new block appeared above you. It could be an AI Overview, but it could equally be a competitor's featured snippet, a video pack or a product carousel. The signal tells you the SERP changed. It does not tell you what changed or whether you are part of it. How to read that click drop with a flat position, and what to do about it, is a topic of its own and not one we will develop here.

The operating conclusion is simple: Search Console remains essential for knowing which queries bring you people and which ones are cooling off, and if you have not wired it into your tracker yet, start with the guide to connecting Google Search Console to a rank tracker. But to answer “am I cited or not?” somebody has to look at the results page. There is no shortcut. What changes between the three methods below is who does the looking: you, an extension, or a data provider querying it for you every day. The full picture of what can be measured continuously is on the Rankiamos feature page.

Being cited is not the same as ranking: the AI Overview citation as its own metric

For twenty years SEO had a single measurement question: what position am I in. The entire instrument set of the industry — trackers, reports, trend charts — was built around that one number. AI Overview citations break the comfort because they introduce a second question the same number cannot answer: am I inside the block Google places above everything else, or not.

They are two independent axes, and seeing them as a four-cell matrix clears the head considerably. You can be high and cited, which is the best case. You can be high and not cited, which is the most frustrating case and gets its own section further down. You can be low and cited, which happens more than people assume and turns a position-14 page into a visible source without it climbing a single spot. And you can be low and not cited, which is where nearly everyone starts.

One nuance to lock in before moving on: the AI Overview does not appear on every query. On brand searches, navigational queries and plenty of simple transactional ones, Google still returns the familiar list. That is why the first thing you need to know about a keyword is not “am I cited” but “does this query show an AI summary at all?”. If the answer is no, the citation metric does not apply and chasing it is wasted effort. If the answer is sometimes, that is already information: your category is in transition and deserves a tighter measurement cadence.

And a scope warning, because it affects what you can promise a client: this is about the AI summary inside Google. It does not include mentions in ChatGPT, Perplexity or any other assistant. Those are different surfaces, queried differently, and they need purpose-built tools. Rankiamos does not track them and we are not going to imply otherwise.

The “rank top 10 and you will be cited” advice no longer holds

When the feature was still an experiment called SGE, the industry landed quickly on a reassuring conclusion: the AI cites whoever already ranks, so keep doing what you were doing. Early analyses backed it up. Around 76% of the links appearing inside the summary came from pages already in the top 10 organic results for that same query. If that were still true today this article would be unnecessary — you could just look at your position.

Analyses published through 2025, with the feature deployed across more countries and languages, find a far smaller overlap, in the region of 38%. In other words, more than half the sources cited in a summary are not on page one for that query. Before anyone builds a strategy on that number, two caveats that need stating and rarely get stated: these are industry studies with different samples, markets and methodologies, and none of them measures Spanish-language queries in Latin American markets specifically. Read them as an order of magnitude and a direction of travel, not as a universal constant that transfers to your niche.

What is solid, and what earns this section its place, is the structural point: the AI summary's source selection is not a copy of the organic ranking. It is a separate selection, made with criteria that overlap without coinciding. And anything that does not coincide with a metric you already have is, by definition, a metric you are missing.

There is an uncomfortable corollary worth accepting early. If you are sitting at position 12 and have spent months fighting your way into the top 10 — the scenario covered in the guide on moving from positions 11-20 onto page one — that work no longer automatically buys the visibility it bought in 2022. Climbing is still worth it, but you are now playing two games and you were only keeping score in one.

There are only three ways to track AI Overviews, and choosing between them is not ideological: it depends on how many keywords you have, whether you need history, and whether anyone besides you will read the data. One fourth option deserves an upfront dismissal, because it comes up in forums often: building your own Google scraper. Technically possible; in practice you inherit blocks, captchas, IP-personalized results and permanent maintenance. Unless you have a data team, it is the most expensive route wearing a free costume.

One caveat that applies to all three: the surface you are measuring depends on the country and language of the query. The same keyword in Colombia and in Mexico can show a summary in one market and not the other, and cite different sources when it does. If you work Spanish-language SEO across Latin America and you are measuring with a setup built for the US or Spain, your data is not wrong — it belongs to a different market. Rankiamos covers the 17 Spanish-speaking countries of Latin America plus Spain and the United States (measured in English), with city-level tracking in 52 cities across 19 countries, and every keyword-plus-city combination consumes one unit of your plan quota.

The free manual method: logging your AI Overview citations week by week

If you are starting out, this is the method to start with, even if you eventually abandon it. It costs nothing, it teaches you more about your SERP than any dashboard, and it leaves you with a history that is worth real money three months in. The whole trick is doing it the same way every time: a sloppily executed manual method produces data that cannot be compared with itself, which is the same as having no data.

Seven steps. The first pass takes a solid half hour; from the third week on, about fifteen minutes for ten queries.

1. Step 1 · Pick 5 to 10 informational queries

   Leave out brand and navigational searches: the AI summary almost never fires there and you will only fill the sheet with zeroes. Prioritize informational, longer, question-shaped queries, which are the ones most likely to trigger the block. If you are unsure, start with the three queries your content answers best and the two where not appearing hurts most.

2. Step 2 · Pin the search conditions and never change them

   Write down the country, language and device you will measure with, and use the same ones every time. If week one is your laptop in Bogotá and week three is your phone on a different network, you are not comparing weeks, you are comparing scenarios. This is the detail that quietly ruins most manual logs.

3. Step 3 · Search in a clean window

   Private window, signed out. It does not remove personalization — your IP still says where you are and Google still adapts results — but it strips the thickest layer, which is your history. Treat what you see as a reasonable approximation, not the absolute truth of the SERP.

4. Step 4 · Log four columns, not one

   Your sheet needs the date, whether the summary appeared (yes/no), whether your domain is among the sources (yes/no), and which domains are cited. That fourth column is the one that later produces the diagnosis: without knowing who Google is citing instead of you, knowing that you are not cited does not help you decide anything.

5. Step 5 · Log your organic position for that week too

   This is the column that makes the sheet useful. With position sitting next to citation you can build the matrix described earlier and see which cell each query falls into. Without it you have a list of yeses and noes with no context.

6. Step 6 · Save a screenshot of the block

   One image per query per week. It costs almost nothing and saves you from two recurring arguments: the client who swears it was showing last week, and the one you will have with yourself in two months when you cannot remember what that summary looked like. Ten consecutive screenshots of the same query also teach you more about what Google picks than ten articles on the subject.

7. Step 7 · Repeat on the same day of the week

   Choose a day and a time slot and stick to them. AI Overview presence is not stable, and if you measure Tuesday mornings and then a Sunday night, you will credit Google with a change that came from your calendar. Four identical measurements a month beat twelve disorderly ones.

Four or five weeks in you will have something almost nobody in your market has: your own series. That is where the method starts paying off. You will be able to say things like “this query shows a summary 3 weeks out of 4”, “these two domains are always cited” or “we dropped out mid-month”. None of those sentences can be pulled from Search Console.

And that is also where the method hits its ceiling. Past twenty or thirty queries, weekly logging stops being sustainable, and with four data points a month you cannot separate a real change from a coincidence on the day you measured. Once you get there the question is no longer whether to automate but what automating costs — which is exactly what comes next. Each plan's quota limits are on the pricing page.

What measuring AI visibility costs today: from 99 USD a month to included in the plan

The market for tools that measure AI visibility assembled itself in under two years and still carries new-category pricing. What matters is less the number than the model: almost every option sells AI visibility as something that gets added to what you already pay, not as something that ships with your rank tracking. That distinction is what decides your invoice.

Third-party prices go stale quickly, so the table is dated. As of July 2026, this is the landscape.

Two warnings about this table. First: third-party prices are list prices and they move, so verify them before making a budget decision, especially in a category that has been repricing itself for two years. Second: you are not comparing equivalent products. A large suite does a great many things Rankiamos does not — technical auditing at scale, backlinks, discovering a competitor's whole keyword universe — and if you need those, a 99 USD add-on can be perfectly reasonable inside a budget that already carries the suite.

The honest comparison is a different one, and it is the one that matters to most independent SEOs and small agencies: if what you need is to know what position in Google you hold and whether the AI summary cites you, does it make sense to pay for a full suite plus an add-on to get those two things? For a freelancer with three clients or an ecommerce site with a hundred keywords, the answer is usually no. That is the gap the Pro plan lives in, and it is why AI Overview citations are not sold behind a separate add-on.

It is also worth naming what no price changes. None of these tools guarantees you a place in the summary, and none of them measures anything outside Google. What you buy with any of them is the data and its continuity over time, which is precisely what the manual method cannot give you once you grow.

This is the situation that brings most people to search for this topic in the first place. The page has held position 2 or 3 for months, the content is good, and the AI block still cites three domains that are not yours. The usual reaction is to start changing things on the page. The correct reaction is to open the log and answer five questions before touching anything.

Notice that none of the five questions is answered by looking at your page. They are answered by looking at the log, which is why the order matters: data first, hypothesis second, change last. Doing it backwards — changing first and measuring afterwards — is the most common way to burn two months and learn nothing, because once the result moves you can no longer attribute it to anything specific.

One point of honesty belongs here. With the data in hand you will be able to say precisely what is happening and since when, which is already far more than most people have. What no log gives you is control: the summary's source selection belongs to Google and there is no lever that guarantees it. Anyone selling you a guaranteed seat in the AI Overview is selling something they cannot deliver.

As for what to do afterwards, one paragraph only, because this article is about measurement and not optimization: the practices the industry associates with getting cited run along the lines of answering the question directly and early on the page, structuring content into short blocks that each resolve one sub-question, being specific with your own data and dates, and keeping authorship and freshness signals in order. None of that is a trick and none of it is a guarantee, and all of it deserves its own article. What we were after here was different: that you stop having an opinion about whether you are cited and start knowing.