Definition
Answer position is the place a brand holds in the order of brands named in an AI-generated answer, counted from 1 at the first brand the answer discusses. If an answer to 'best CRM for startups' covers HubSpot, then Pipedrive, then your product, your answer position in that response is 3.
The metric exists because a mention is not a single outcome. Being the opening recommendation, the fourth row of a comparison table and a one-line alternative in the last paragraph all count as 'mentioned', yet they lead to very different results. Readers of AI answers tend to act on the first concrete recommendation, so position captures what mention rate and share of voice cannot.
Answer position is the closest AI search equivalent to rank in SEO, with important differences. There is no fixed results page. The same prompt can produce a different set of brands, in a different order, on the next run, on another model, or for a user in another country. So position has to be measured across many responses and reported as an average or distribution, never read from a single answer (see AI answer volatility).
In Promptwatch, position is calculated by reading each response top to bottom, across prose, headings, lists and tables, and assigning each substantively discussed brand a position from its first qualifying appearance. Positions count brands, not words, and run 1, 2, 3 without gaps. A name that appears only in a cited URL or source title does not establish a position. The reported value is the average across responses that mention the brand, so '#1.4' means the brand is usually named first. Responses that leave the brand out are not counted against its position; that gap shows up in coverage metrics instead. Position also feeds into the AI visibility score, which adds share of attention and depth on top of placement.
Some teams prefer bands over an exact average: first mention, top three, fourth or later. Bands are easier to explain to executives and less sensitive to answers that list ten or more brands. Whichever method you use, keep it fixed so that trends are comparable over time.
Position moves for reasons outside your own site. Models lead with the brand they most strongly associate with the prompt's exact wording, so niche prompts ('best CRM for dental practices') often show a better position than broad ones. Broad third-party coverage that pairs your brand with the category ('Acme, the project management tool for agencies') tends to pull it up the order. And your position can fall with nothing changing on your side because a competitor became the default first answer. Slice by model, prompt and topic before drawing conclusions, and compare your average with competitors' in a competitor analysis. Our guide to AI search visibility KPIs covers how to report position alongside presence and sentiment.
Examples of Answer Position
- A project-management brand appears in 60% of tracked ChatGPT answers but averages position #4.2, while a competitor appears in 45% and averages #1.3, which shows the brand has a prominence problem rather than a coverage problem.
- A CRM vendor holds position #1 on prompts about dental practices but #5 on 'best CRM for small business', and decides to invest in third-party roundups that cover the broad category.
- An agency reports answer position in three bands (first, top three, later) for a client's board deck because the exact average swung by half a position from week to week.
- A brand's average position on Perplexity drops from #2 to #3 in one week; a check of competitor positions shows a rival launched a widely covered product, not that anything changed on the brand's site.
