Keyword Rank Checker and AI Visibility Tracking

Rank tracking measures only one of the surfaces where a company now gets chosen or skipped. What can be measured on Google and in each AI assistant, how to start for free, and why two tools never report the same numbers.

Published 30 September 2026 — Co-founder of DoGeo, responsible for the technical side of GEO

Short answer

The free and most reliable keyword rank checker for Google is the Performance report in Google Search Console, which gives the average position of every query and page from Google’s own data. A rank tracker adds daily tracking of a chosen keyword list and of competitors. Neither tells a company whether ChatGPT, Perplexity, Gemini or Claude name it: that is measured with a fixed set of prompts and a log of whether the brand appears, with which source and where in the answer.

What is a rank tracker

Definition

A rank tracker is a tool that queries a search engine on a schedule for a list of keywords and records the position each website holds. It is used to follow the same searches over time, your own and your competitors’.

A rank tracker simulates the search: it runs it from a fixed location, language and device type, and stores the results page it sees at that moment. That is why it gives an exact position per keyword, but it is the position of one sample, not the one every searcher saw.

Search Console, by contrast, averages the real impressions Google recorded, so the two sources do not give the same number.

How to check your Google rankings for free

Search Console

The free way to check keyword rankings on Google is the Performance report in Google Search Console, with the “Web” search type. It shows the average position of each query and page based on the real impressions Google recorded.

Three steps to read it properly:

  1. Open Performance, choose the “Web” search type and turn on average position next to clicks and impressions.
  2. Filter by one query or one page, because the site-wide average mixes searches that have nothing to do with each other.
  3. Compare two periods of the same length, not one day against another.

Search Console Help defines the value precisely: for each query it counts the topmost position a link to the site held, then averages those positions across every query where the site appeared. If a search returns the site at positions 2, 4 and 6, that search counts as 2.

Searching by hand is no substitute for the report: according to the same Google Help page, search history and location change the position.

Why average position loses meaning when a model writes the answer

Limit

Average position was designed for a list of links, and generated answers are not a list. In Google’s AI Overviews every cited link gets the same position, so the number cannot tell the main source from the last one.

Google documents that sites appearing in AI Overviews and AI Mode are included in the overall search traffic in Search Console, under the “Web” search type (Google Search Central, “AI features and your website”, updated 10 December 2025). Search Console Help adds that an AI Overview occupies a single position in the results and that “all links in the AI Overview are assigned that same position”.

Nor can one be inferred from the other. In an analysis of 15,000 prompts published by Ahrefs on 11 August 2025 and updated on 31 May 2026, on average only 12% of the links cited by the assistants studied appeared in Google’s top 10 for the same query. Why that happens, and what to do about it, is covered in why ranking on Google is not enough for ChatGPT.

What you can measure on Google, ChatGPT, Perplexity, Gemini and Claude in 2026

Comparison

In 2026, of the engines customers use today, only Google and Bing document a report for websites on their presence in AI answers. In ChatGPT, Perplexity, Claude and the Gemini app, measurement comes from their robots’ visits in the site’s firewall and, from the outside, from asking and logging.

SurfaceReport from the providerWhat that report measuresWhat is left to measure from outside
Google’s classic resultsGoogle Search Console, Performance reportClicks, impressions and average position from real dataDaily positions for specific keywords and competitors
Google’s AI Overviews and AI ModeGoogle Search Console, inside the “Web” search typeClicks and impressions; every link in an AI Overview shares one positionWhether the brand is named in the text, and where
Bing’s AI answersBing Webmaster Tools, AI Performance report (preview)Citations, cited pages and, since June 2026, citation share per queryThe role each page plays inside an answer
ChatGPT, Perplexity, Claude and the Gemini appNone documented by the providerNothing; their robots do show up in the site’s firewallWhether the brand appears, with which source and where in the answer
Sources: Google Search Console Help and Google Search Central, consulted on 30 September 2026; Bing Webmaster Tools blog, 10 February and 16 June 2026.

What is share of citations and how is it calculated

Metric

Share of citations is the part of the citations in a set of answers that goes to one brand: the times its site is cited divided by all citations logged in those same answers. It replaces position where there is no list.

Bing Webmaster Tools added it as a preview on 16 June 2026 under the name Citation Share, defined as how much of the citation space a site receives for a specific grounding query. Bing’s own blog states that it is an observational metric, not a ranking system.

Outside Bing, share of citations is calculated from your own log: the same prompts, the same engines and the same dates, counting the brand’s citations and everyone else’s. Mention, placement in the text and accuracy of what is said are the other metrics, covered in how to tell whether AI citations are working.

How to track for free whether ChatGPT, Perplexity, Gemini and Claude cite you

Manual method

The free method to track whether AI assistants cite a company takes three steps: a fixed list of real customer prompts, a fixed frequency and a log with the same fields every time. A spreadsheet is enough.

  1. 01 / PROMPTS

    Fix the questions a customer would ask

    Real buying questions for the category, written the way someone who does not yet know the brand would write them, and without naming it. The list is frozen: if it changes every month, the series stops being comparable. If the brand never appears, it is worth first checking whether AI crawlers can read the site.

  2. 02 / FREQUENCY

    Run them the same way, several times

    Same day, a fresh session with no history and web search on where the assistant allows it. Each prompt is run several times, because an assistant almost never returns the same list of brands twice.

  3. 03 / LOG

    Record whether you appear, with which source and where

    One row per prompt, engine, run and date, with four fields: whether the brand appears, which URL is cited as the source (your own or a third party’s), where in the answer it sits and which competitors are named. Keep the full answer as well.

What types of rank tracking tools exist and what each one measures

Tools

Tracking tools fall into six categories depending on where the data comes from: the search engine itself, scheduled queries, estimates, prompts sent to assistants, their robots’ visits or a manual log. No category covers every surface.

CategoryWhere the data comes fromWhat it measures wellWhat it does not measure
Search engine consoles (Search Console, Bing Webmaster Tools)Real data from the search engine itselfClicks, impressions, average position and, on Bing, AI citationsAnything outside its own search engine and your own site
Classic rank trackersSimulated searches from a fixed location and deviceExact position per keyword, day by day, competitors includedWhat each real person sees, or whether an assistant names the brand
Traffic estimators in SEO suitesEstimated positions multiplied by estimated search volumesComparing orders of magnitude between competitorsThe real visits of a specific site
Prompt tracking in AI assistantsPrompts sent on a schedule to several assistantsMentions, cited sources and share of citations at scaleA stable answer: each run can change the list
Robot logs at the firewall or CDNVisits from each assistant’s robots, by name and IPWhich pages each assistant reads and whenWhat the assistant then writes in the answer
Manual logPrompts run by hand and logged in a spreadsheetThe same as prompt tracking, with fewer prompts and no costEnough volume to detect small changes

It is the only first-party source on ChatGPT, Perplexity or Claude that does not depend on asking them: the firewall records every visit from their robots, including those that open a page at someone’s request. This is what DoGeo sets up for each client, with a monthly count by robot and by page.

Why ranking and visibility data never match across tools

Discrepancies

Data never matches across tools because each one measures something different: Search Console averages real impressions, a rank tracker takes a sample and an SEO suite estimates. Assistants add one more cause: the same prompt almost never produces the same answer.

Estimates miss in both directions. Comparing DataForSEO’s estimate with Search Console on three sites, DoGeo found that for Kulturo the monthly organic visits estimated from Spain alone were 2.7 times the real clicks from all countries in August 2026, and for HeyTalent 3.4 times. On a site specialised in rigid inflatable boats the estimate fell short: 6 estimated visits a month against 108 real clicks from Spain in seven weeks.

In the first two cases the comparison favours the tool, which estimates one country against clicks from all of them, and it still overestimates. These are three sites: enough to distrust an estimate, not enough to set a margin of error.

In AI assistants the variation is larger. SparkToro, with Gumshoe, had 600 volunteers run 12 prompts through ChatGPT, Claude and Google’s AI a combined 2,961 times, and Rand Fishkin concluded on 28 January 2026 that there is less than a 1 in 100 chance that ChatGPT or Google’s AI will give the same list of brands in any two responses.

What AI presence figures Search Console and Bing Webmaster Tools give today

Data

Both consoles already report AI presence figures that no rank tracker sees. Kulturo logged 8,700 AI citations in three months, with an average of 10 cited pages, according to Kulturo’s AI performance report in Bing Webmaster Tools.

HeyTalent logged 14,200 AI citations in three months, with an average of 19 cited pages, according to HeyTalent’s AI performance report in Bing Webmaster Tools.

On Google, HeyTalent went from 1,485 impressions in the first week of its generative AI features report, 15 to 21 June 2026, to 9,500 in the week of 8 to 14 September: 6.4 times as many, according to HeyTalent’s Google Search Console. These are impressions, not clicks or citations, and the report is in beta; how to read it is in how to tell whether Google already cites you.

Measuring it has commercial consequences: between 15% and 20% of new customers at Kulturo and HeyTalent come through GEO, according to what they say when they sign, and that route does not show up in average position.

Keep reading

Let’s talk about your visibility

Book a call with DoGeo. We come to it with a report already prepared on how Google and AI assistants see you today, and on what we would do first.