Generated from the Orbator AI Recommendation Index — 143424 citations across 271 categories, trailing 28 days. Licensed CC BY 4.0 with attribution to orbator.io.

The headline

When AI assistants (ChatGPT, Claude, Gemini, Perplexity, Grok) recommend software, 46.2% of the sources they cite are owned by a vendor whose product appears in the recommendation index — comparison pages, "best X" listicles, and category guides published by companies that sell in that category.

AI answers are not neutral syntheses of independent reviews. To a measurable degree, they are recycling vendor marketing back to buyers as advice.

How we measured it

We run 143424+ live buyer-style prompts per 28-day window across five AI engines and record every source each engine cites before answering. Each cited domain is matched against our registry of software vendors (canonical domains, 46.2% match = the citation resolves to a company that sells a product our Index has seen recommended). Search-infrastructure noise (e.g. vertexaisearch redirects) is excluded. Full method, sample sizes, and confidence intervals: https://www.orbator.io/ai-index/methodology — raw stats at https://www.orbator.io/data.

Why it matters

1. For buyers: "the AI recommended it" increasingly means "a vendor wrote the page the AI read."

2. For challenger vendors: the winning move is not more ads — it is becoming a source AI engines actually cite.

3. For the ecosystem: independent, measured recommendation data is scarce — which is exactly why we publish ours openly (CC BY 4.0).

What is in the open data

Per-category recommendation shares with 95% Wilson confidence intervals, per-engine agreement, first-recommendation rates, and the live model versions measured — updated weekly at https://www.orbator.io/data.

Questions or want the underlying rows? hello@orbator.io — attribution: orbator.io.