AI search rankings

An AI search ranking is a per-brand score for a single category, computed by running a fixed set of buyer-intent queries against one or more AI engines, capturing the answers, and counting two things: the percentage of answers that name the brand in the prose (visibility), and the percentage that cite the brand's domain in the source list (citation rate). A composite of both, weighted, becomes the rank.

The ranking is reproducible because the queries are fixed (the same 8 or 40 questions every cycle), the engine is fixed (e.g. Google AI Mode), and the scoring rules are fixed (substring match for name, apex-domain match for citation). Two runs a month apart can be compared directly. That makes AI search rankings the closest analogue to classical SERP rank tracking, just applied to a different surface.

Three structural differences. First, the surface is different: AI engines synthesise an answer rather than return a list, so 'rank' here means 'how often you're in the answer at all' rather than 'are you position 3 or position 8'. Second, the volatility is higher: AI engines regenerate answers on every query, so the same question 10 minutes apart can return slightly different prose and a different citation list. You smooth this with repeated sampling. Third, the levers are different: domain authority matters less, on-page structure (FAQ schema, server-rendered HTML, llms.txt) matters more.

For most brands the right move is to track both: classical SERPs for branded and high-commercial-intent queries, AI search rankings for the same categories. They're correlated but not identical, and movement on one does not always mean movement on the other.

A useful ranking exposes the inputs as well as the score. At minimum: the engine that produced it, the date of the last refresh, the query set (so a sceptic can verify), the per-brand visibility and citation numbers, and the list of cited source domains across the category. Without those, the ranking is a black box and a buyer cannot tell whether they were measured fairly or whether the data is fresh.

monitoraeo's public industry rankings publish all of this per category, refreshed monthly via Google AI Mode (SearchApi.io). Paid audits run a 40-query brand-aware set across 5 engines and produce a per-brand ranking plus a fix list.

Two leverage points, ordered by impact. First, fix the technical layer (GEO): allow GPTBot, ClaudeBot, PerplexityBot, and Google-Extended in robots.txt; render content server-side; publish FAQPage and Article JSON-LD; publish a /llms.txt. None of these are optional. Without them you are invisible to the retrieval step that happens before the AI writes its answer.

Second, work on the content + entity layer (AEO): write answer-shaped 2 to 3 sentence chunks for the queries you want to win, earn presence on the third-party domains the AI already cites in your category (typically Reddit, G2, Capterra, Wikipedia, trade pubs), and tighten your Organization schema with sameAs links to Wikidata + Crunchbase + LinkedIn. Run an audit first so you know which gap is binding.