Average position: definition, and why AI answers have none

Average position is the mean rank at which your highest result appeared in a search engine's results, averaged over every impression it earned. Lower is better: an average of 1.0 means you were always the top result shown.

It is an average, so it hides spread. Say your page shows at position 1 for one search and position 9 for another. The report says 5, a position the page never held.

The metric depends on one assumption: results arrive as an ordered list of links, and each link has a slot. Remove the list and the number has nothing to measure.

Google Search Console shows average position next to impressions, clicks and click-through rate in its performance report. The same dimensions are available through the Search Analytics API, filtered by query, page, country or device.

Generative answers break the list assumption. An AI Overview is a written answer with supporting links, not ten ranked results. Google's own documentation says AI Overviews and AI Mode "may use different models and techniques" and can return different links. A page can be quoted in the middle of a paragraph, listed last, or used without a link at all.

For how classic rank still connects to AI citations, see organic ranking. Ahrefs found that 37.9% of URLs cited in AI Overviews also appeared within the first 10 results (March 2026). That is an association, not a path you can engineer.

As of October 2026, Google's Search Console report for generative AI features exposes impressions only. It has no clicks, no CTR and no position column. Google's page on third-party SEO tools adds that "Third-party tools don't have access to our internal ranking data" and "They can't guarantee performance."

Other engines disclose even less. OpenAI says ChatGPT search rewrites your prompt into one or more targeted queries, and it decides per prompt whether to search at all. Nobody outside the engine sees those queries.

What this means: a tool that reports "your AI rank is 3" is reporting its own construct. That may be useful internally, but there is no ground truth to validate it. Our guides on AI rank checkers and why position isn't the metric cover this in depth.

The next move is to track things you can verify:

• Whether each engine names your brand for a given buyer question.
• Whether it cites your pages, and which ones.
• Whether the facts it states about you are right.
• Whether the answer used live search or came from the model's training data.

The free checker on this site gives you a snapshot of the first three for up to three buyer questions.

• "We rank number 1 in ChatGPT." No engine publishes a position for generative answers. Treat the claim as a vendor's model, not a measurement.
• "The first citation is the top rank." Citation order is not a published ranking. Mechanistic research (arXiv:2606.28358) repaired over 90% of missed citations by adjusting internal model components without changing the sources. Citation is a noisy proxy for influence. Read more on where your brand appears in an answer.
• "Context position is a lever." C-SEO Bench found that a document's position inside the model's context mattered more than rewriting tactics. That position is set by the retrieval operator, and the study placed documents there by hand. See position bias.
• "One average across engines tells the story." Engines use different indexes, models and citation formats. A blended number hides which one is failing, as our guide to AI visibility scores explains.

• Organic ranking: classic rank and its association with AI citations.
• Position bias: why order inside a model's context affects what it cites.
• Position-Adjusted Word Count: the lab metric behind the widely quoted GEO uplift figures, measured on content already in context.
• AI visibility scores: what blended scores average and what they leave out.

Is average position still useful for SEO? Yes, for classic search results. It shows roughly where your pages sit for the queries that triggered them, and changes over time can flag a ranking drop. Read it alongside impressions and clicks, because an average hides spread. It does not describe how you appear inside AI Overviews, AI Mode, ChatGPT or any other generative answer.

Does Search Console show average position for AI Overviews? No. As of October 2026, the generative AI report in Search Console shows impressions for AI Overviews and AI Mode only. There are no clicks, no CTR and no position. Google also had a confirmed bug that depressed impressions from around 12 August 2026, so check data from that window before trusting a trend.

What should I track instead of an AI rank? Track per engine: whether your brand is named for real buyer questions, which of your pages get cited, and whether the stated facts are correct. Record whether each answer used live search, because answers without retrieval come from training data with a fixed cutoff. Keep engines separate rather than averaging them into one number.

Can I improve my position inside an AI answer? Nobody can promise that, because no published position exists to move. What you control is access: content in the HTML before JavaScript runs, robots.txt that allows each engine's search crawler, a firewall that doesn't block them, and index eligibility. Those remove blocks. Which retrieved pages the model then uses is decided by the model.