AEO · 7 min read

Is rank tracking still enough? How to measure your answer share in AI search

Accendoz·
You can rank first and miss the answer. So track answer share too.

Not on its own. Keep your rank reports, because Google still sends far more visits to websites than ChatGPT, but add answer share: how often AI answers to your buyers' questions name your business. A ranking cannot tell you that, since most AI Overview citations now come from outside the top 10.

No, not on its own. Keep tracking rankings, because Google still sends far more visits to websites than ChatGPT does, but add a second number: answer share, the percentage of AI answers to your buyers' questions that name your business. A ranking can no longer tell you whether you were named, because in an Ahrefs study published in March 2026, only about 38% of pages cited in AI Overviews ranked in the top 10 for the query.

Answer share is simply share of voice applied to AI answers. The hard part is measuring it properly: AI Overviews, for one, change their content about every two days on average, so a single screenshot proves very little. Here is why the old report falls short, how to score answer share, and a monthly routine a small team can keep.

of visits to a page with an AI summary included a click on a link inside it (Pew, US)
1%
of visits to a page with an AI summary included a click on a link inside it (Pew, US)
of AI Overview citations ranked in the top 10 for the query (Ahrefs, March 2026)
38%
of AI Overview citations ranked in the top 10 for the query (Ahrefs, March 2026)
average life of an AI Overview before its content changed (Ahrefs)
About 2 days
average life of an AI Overview before its content changed (Ahrefs)
more traffic Google sent to websites than ChatGPT (Ahrefs, February 2026)
190x
more traffic Google sent to websites than ChatGPT (Ahrefs, February 2026)

01

Why a ranking no longer tells you enough

Rank tracking assumes the searcher sees a list and clicks something on it. AI answers break both halves. In Pew Research Center's study of real browsing by 900 US adults, people who saw an AI summary clicked a traditional result on 8% of visits, against 15% without one, and clicked a link inside the summary on just 1%.

The position figures in your reports have changed meaning too. Google's Search Console help says every link inside an AI Overview is assigned the same single position, and a follow-up question in AI Mode counts as a new query. Google also reports AI feature traffic inside the ordinary Web search data. A page can sit at position one and still be missing from the answer the buyer actually read.

And position one is worth less where an AI Overview appears. Ahrefs, which sells SEO software, reported in February 2026 that by December 2025 an AI Overview correlated with a 58% lower average desktop click-through rate for the top-ranking page.

02

Ranking and being cited have come apart

If AI answers simply quoted the top results, rankings would still be a good proxy. They do not. Ahrefs found that the share of AI Overview citations ranking in the top 10 fell from about 76% in its July 2025 study to about 38% in its March 2026 update. Ahrefs had also improved how it collects citations in between, but it links the shift to Google running more related searches behind each query and drawing sources from those.

Google's two AI surfaces also disagree with each other. For the same US queries, AI Mode and AI Overviews cited the same URLs only 13.7% of the time. A business can be named in one surface and missing from the next on the same day. One rank report cannot show that. An answer share check can.

03

What answer share is, and how to score it

Answer share is the percentage of AI answers to a fixed set of buying questions that name your business. Pick the 20 or so questions a customer asks before they buy, run each one on the assistants your customers use, and record what comes back.

The arithmetic is simple. Take a hypothetical physiotherapy clinic in Kozhikode that asks 20 questions on four assistants, three times each: that is 240 answers. If it is named in 30 of them, its answer share is 12.5%. Run the same count for its three nearest competitors and it knows where it stands. For every answer, record five things:

What to record for every AI answer
FieldWhat you write downWhy it matters
NamedIs your business mentioned at all?The core of answer share
LinkedIs one of your pages cited as a source?The route to a visit
OrderFirst, second or later among the businesses namedThe recommendation or an also-ran
AccuracyAre your hours, location, services and prices right?Wrong facts cost enquiries even when you are named
CompetitorsWhich other businesses are namedTurns your score into share of voice

04

Answers move, so sample them

The biggest mistake is treating one answer as the truth. Ahrefs tracked more than 43,000 keywords for a month and found that AI Overview content changed roughly every two days. Between consecutive versions, only 54.5% of cited URLs stayed the same on average, and about 46% of the named brands, people and places changed.

The meaning stayed far steadier than the wording, which suggests the topic you are trusted on matters more than any single phrasing. In practice: ask each question several times, keep the wording and conditions the same each month, and judge trends over three months rather than reacting to one week.

Your brand can be cited today, and gone tomorrow.

Louise Linehan, Ahrefs, November 2025

05

Turn every gap into a task

Each question where you are absent becomes a piece of work. Usually it is one of three things: you have no page that answers the question directly, your page answers it but buries the answer, or other sites describe your category and never mention you.

Accuracy deserves the same attention as presence. In an Ahrefs experiment with an invented brand, Gemini and Perplexity repeated detailed false claims planted on third-party sites in roughly 37% to 39% of answers, while ChatGPT, which often cited the brand's official FAQ, stayed under 7%. Most of the assistants tested still trusted a fake investigation over the official FAQ, so a clear, specific page of your own facts is no guarantee. It does give the assistants that check sources something accurate to cite.

Mentions elsewhere count as well. In Ahrefs' study of 75,000 brands, mentions of a brand across the web, and on YouTube in particular, correlated far more strongly with AI visibility than backlinks did. Correlation is not proof, but it is a sensible place to aim.

06

A monthly routine that fits a small team

  1. Fix the question list. Write down 20 questions buyers ask before choosing a business like yours, in their words, including the city or locality, and in Malayalam too if that is how your customers ask.
  2. Choose the assistants your customers actually use, for example Google's AI Overviews and AI Mode, ChatGPT, Gemini and Perplexity.
  3. Run each question three times per assistant in the same week, logged out where possible, and fill in the record for every answer.
  4. Score yourself and three competitors, and chart the trend next to your rankings and your enquiries.
  5. Pick the three biggest gaps and turn each into a fix: a new answer page, a rewritten opening paragraph or a corrected directory listing.
  6. Recheck next month and keep a change log, so you can see which fixes moved the number.

Rankings still belong in the report: in Ahrefs' traffic data, Google sent about 190 times more visits to websites than ChatGPT. But they now sit beside answer share rather than above it. If you would like help setting this up for your category, talk to our SEO and AEO team.

Sources

Facts checked against the sources below and current as of 8 May 2026. Several figures come from Ahrefs, an SEO software company that publishes its own research.

Chat with us