What replaces rank tracking when an AI gives the answer?

Direct answer

Rank tracking measures a page's position for a query a person typed, and generative engines rewrite that query before searching. What takes its place is the brand's presence in the answer and the sources that travel with it, inside one model and one window. In the benchmark, the same e-commerce category had three different leading brands spread across 21 markets, a difference a single position has no room for.

What stops working

The query reaching the search index stops being the one the person wrote. Google describes AI Mode issuing several related searches concurrently across subtopics and bringing those results together into one response¹. Microsoft documents that Microsoft 365 Copilot Chat generates a short query of a few words and sends it to Bing, different from the original prompt². Google documents that the Gemini API with search generates one or more queries before answering³.

With the query rewritten by the model, a position for a word chosen in advance loses what it referred to. Position tracking still describes classic search well; what it loses is the ability to describe the generative surface.

There is a second loss, on the result side. OpenAI documents that its web search tool returns a field with every URL retrieved, wider than the citations that end up displayed⁴. A domain can enter the answer's context without appearing in any visible citation, so what shows on screen is a subset of what the engine used.

What gets measured instead

The unit becomes the brand's presence in the answer, measured over a repeated bank of prompts. Three numbers replace the position: in how many answers the brand appears, what share of the category's mentions it holds, and which sources travel with those answers.

That way of measuring has backing on the provider side. Bing Webmaster Tools reports to a site owner the total citations, the daily average of cited pages, the grounding queries the AI used to retrieve the content, and citation activity per URL⁵. Grounding queries are exactly the piece position tracking takes as given and a generative engine builds on its own.

Research adds a distinction worth carrying into the measurement. A 2026 work separates a document's contribution to the answer from its citation, and observes that what sends traffic back to the creator is the citation⁶. Measuring influence over the text alone overstates the result; measuring visible citations alone understates the use.

The Pew Research Center, a research centre with no SEO product, measured across 68,879 searches by 900 US adults in March 2025 that 88% of AI summaries cited three sources or more⁷. An answer citing several sources at once has no first place to win.

A concrete example

In the benchmark, across 1,150 answers in the e-commerce platforms category over 21 markets, Shopify led 18 of those markets and the other three went to Magento and Tiendanube. The brand is the same and the category is the same; what changes is the market. A single position per searched word has nowhere to keep that difference, and a share of voice per market does.

The second example holds the country fixed and changes the language. Across 108 answers in the payments and fintech platforms category, Mercado Pago held 3.0% of mentions in Spanish and had no mentions in English, in the same window and with the same bank of prompts. The difference between being there and being absent fell on the side of the prompt language.

Both cases read the same way: presence depends on the cut, and the cut is chosen when measuring. A figure with no market, no language, no model and no window describes less than it appears to.

How I measure it

I write a bank of unbranded prompts for the category, in the market's language, and I repeat it several times per model. I record per answer which brands get named and which domains get cited, and I keep the collection method next to every figure.

I always compare inside the same model, the same market and the same window. An average across models hides the case of a brand that leads in one and vanishes in another, and a comparison across different windows mixes site changes with engine changes.

On that base I build three series that do follow over time: the proportion of answers naming the brand, its share of the category's mentions, and the domains travelling with those answers. All three compare across weeks while the prompt bank and the collection method stay fixed.

How to start measuring your AI visibility

1. Start with a small bank of questions, five to ten, written the way a customer would ask them and without naming your brand. Unbranded is worth the trouble because that way the model decides who appears and your prompt stays out of the decision.

2. Repeat every question several times on each engine that matters to you. A single run mixes the signal with the noise, and you end up celebrating or mourning one draw.

3. Record three numbers per run: how many answers name your brand, what share of the category's mentions it takes, and which domains travel with it. Those three replace the position, and all three can be followed over time.

4. Always split by market, by language and by model. The two examples above show the leader changing with the country and with the language, so an average hides the very case you need to see.

5. If your site is in Bing Webmaster Tools, look at the AI Performance report before paying for a tool. It is free and it hands back the grounding queries, the piece rank tracking used to take as given.

6. Save the first measurement as a baseline and measure again with the same bank. Keeping it fixed is worth the discipline because changing it halfway stops the series from comparing with itself.

Which AuraMetrics modules address this

Measure the brand with the same category prompts. Prompt Tracking runs prompts on the five engines.

See which sources AI cites when it answers about the category. Prompt Tracking shows, for each prompt and model, the sources the answer cited, and AI Market Leaders groups them by brand.

Measure the traffic that arrives from AI. AI Impact connects to Google Analytics 4 through read-only OAuth and shows traffic and revenue from AI sources.

Frequently asked questions

What alternatives are there to traditional rank tracking for measuring visibility in AI search?

The alternative with backing is presence over a fixed bank of prompts, repeated per model and per window, with the cited sources alongside. A first party already publishes that shape of measurement: Bing Webmaster Tools reports citations per URL and the grounding queries the AI used to retrieve the content⁵.

Is a single AI visibility score across several answer engines any use?

A single score across engines hides the difference that matters most. According to a study by Profound, a company that sells AI visibility analytics, across 680 million citations between August 2024 and June 2025, Wikipedia held 7.8% of ChatGPT's citations while Reddit held 2.2% in AI Overviews⁸. An average of those two describes a surface that has no existence.

What does an AI visibility score platform have to declare?

A score works when it states which models went in, with which prompt bank, in which market and in which window. Without those four things it loses the ability to compare with itself from one week to the next.

How I measured it

AuraMetrics weekly benchmark: a fixed bank of unbranded prompts per category and market, run across several models. The e-commerce case comes from 1,150 answers over 21 markets; the payments and fintech one, from 108 answers with the prompt in Spanish and in English. Both figures are already published in Quick Insights.

Each external claim in this guide points at an entry in the public source registry, with its consultation date.