AttractOS Benchmarks

Who AI assistants actually cite, by category

Each benchmark is a single dated probe run: we query three live assistant-search lanes across a set of real buyer intents and record exactly who they cite, with raw answer excerpts as evidence. These are sampled snapshots, not rank trackers — a dated read of one moment, useful for spotting where a domain clearly is or isn't showing up. Every page carries its full methodology (model IDs, probe date, single-sample caveat).

Categories

Answers — one buyer intent at a time

Each benchmark aggregates five buyer intents. Want the read for a single query — who gets cited when you ask exactly "best free crm for startups," with that query's own raw evidence? That's the atomic unit:

• Browse every per-intent snapshot →

Guides — the capability gap behind these benchmarks

Why do these benchmarks exist at all? Because no assistant can tell you, on its own, who it cites — asking ChatGPT directly just triggers a fresh one-off search, not a report. These guides explain the gap and honestly survey the tools built to close it:

• Can ChatGPT tell you if it recommends your website?
• Best AI visibility monitoring tools — ranked overview, named competitors included
• AttractOS vs the field — head-to-head comparisons with live-checked pricing and rows we concede
• Why your rank tracker can't see ChatGPT
• AI visibility tools vs. traditional rank trackers
• Why doesn't ChatGPT recommend my business? — evidence-based diagnostic
• How to get your business recommended by ChatGPT — honest guide, no guarantees
• What is llms.txt and do you need one?
• Do AI chatbots crawl your website? Which bots, and how to check
• How does Perplexity choose its sources? — from our own probe corpus
• How does Gemini grounding pick citations? — from our own probe corpus
• AEO vs. SEO: what actually changes
• What is generative engine optimization (GEO)?

How to read these

Each cell in a benchmark is a single, dated sample from a live API — not an average over time and not a rank tracker. Assistant answers vary between runs and shift as models and their indexes change. We aggregate ranked counts from the Perplexity Sonar and Gemini grounded lanes, where domain citations are consistently present, and preserve the OpenAI web-search lane's excerpts as evidence even when it names brands without linkable URLs. Full detail lives in our methodology.

Want to see where your own domain lands on these same intents? Every benchmark carries a free, prefilled scan — or start a scan from scratch.