AI search ranking

In classical Google search, ranking is a single position number per URL per query. In AI search, ranking is fuzzier and has two distinct dimensions. Citation rank: when the AI's answer includes a source list, which position is your domain (first cited, third cited, eighth cited). Mention rank: when the AI names brands in the answer prose, which order are they named in (the first-named brand carries the most weight, like the top blue link).

Both rankings can shift between runs of the same query because AI search engines are non-deterministic. So 'ranking' is better measured as a rate (cited X% of the time, named in position 1 X% of the time) rather than a single position number.

Three structural differences. (1) No single ranking algorithm. Each engine (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews) ranks differently. Google AI Overviews leans on Google's classical ranking signal as the input; ChatGPT uses Bing then re-ranks; Perplexity has its own crawler and ranker. Optimising for one does not automatically lift the others. (2) Extractive content beats backlinks. Pages that answer the question directly in 2 to 3 sentence blocks win against pages with stronger backlink profiles but essay-form content. (3) Topical authority more than domain authority. A specialist site on a narrow topic outranks a generalist with higher classical DR.

What still matters: server-rendered HTML, fast load, structured data, allowing AI crawlers in robots.txt. These are baseline requirements, not differentiators.

The same diagnostic pattern works across all five engines. Pick a representative query set (40 to 100 buyer-intent questions in your category). Run each query through each engine, ideally 3 to 5 times to average out the non-determinism. For each run, capture: (a) whether your brand is named in the answer and at what position, (b) whether your domain is cited as a source and at what citation position, (c) which other brands are named or cited.

Aggregate to four numbers per query: mention rate, mention position, citation rate, citation position. Repeat monthly. The month-over-month delta is your real product, single snapshots are hard to act on. Our full audit automates this across five engines on a 40-query buyer-intent set per brand.

Two leverage points, in order of speed. (1) Earn citations on the aggregator sites your category's AI engines already trust: G2, Capterra, Reddit, Wikipedia, Healthline, NerdWallet, depending on industry. Citation rate moves within 2 to 4 weeks of new earned mentions. (2) Restructure your own content to be extractive: 2 to 3 sentence answers to specific buyer questions, FAQPage schema, recent dateModified, server-rendered HTML. Same-page changes move citation rate in 7 to 14 days.

What does not move AI search rankings: keyword density, link-building from low-authority sites, AI-generated marketing fluff (engines actively down-weight this), classical SEO tactics that target Google ten-blue-link ranking signals.