Original research
What this research covers. AEOForged runs live questions to ChatGPT, Gemini, and Perplexity, records every cited source and every miss, and publishes the article, methodology, and dataset. The live corpus spans 96 questions and 3,563 recorded citations across four independent studies. These sweeps do not cover Google AI Overviews.
Each study uses Wilson 95% confidence intervals and a same-day variance panel. Raw datasets are published alongside every article. Numbers are frozen at measurement — nothing is invented after the fact. Purchasable sector editions live on Sector Intelligence; packages that use this measurement stack are on Pricing. Research updated .
Not a frozen sector study — our own domain on the same citation instruments, updated as daily checks land. Also at /research/live-citation-board.
Monthly measured experiment pools. Every figure is measured on the named instrument versions; no client is named. All Citation Lab editions.
No edition has been published yet.
Dated observations from sites we operate. Same instruments, not a sector census. Also at /research/llmstxtgen-chatgpt-referrers.
Product-led guide to purchasable sector artifacts — distinct from first-party dogfood above. Storefront: /intel.
AEOForged Sector Intelligence Reports as purchasable AI sector reports: ChatGPT, Perplexity, and Gemini panels, Wilson CIs, battlegrounds, snapshot vs Dominate.
Read the guide
How a multi-brand sector benchmark differs from your own citation board — frozen panels, misses, confidence intervals, and Sector Intelligence Reports.
Read the methodology explainerSource-list rates by engine, then the July citation-mix rollup those rates sit beside.
ChatGPT returned source URLs in 25 of 96 July AEO-category answers (26%, Wilson 95% CI 18–36%). Gemini and Perplexity sourced every recorded check. Mentions and citations are reported as separate series.
Read the paper
Across four studies — 96 questions, 3,563 recorded citations — editorial content captured 89.5%. Only youtube.com and linkedin.com placed in the top five in every sweep.
Read the July synthesisOne measured paper per topic. Each line: the finding, then article · methodology · dataset.
Reddit led at 30.8% (95% CI 22.2–40.9%). Trust and “is AEO a scam” queries cite community threads more than vendor pages.
Read the articleLinkedIn led at 34.8% (95% CI 25.8–44.9%). Authority-signal queries preferred social corroboration over brand sites.
Read the articleLinkedIn and YouTube tied as leaders at 17.4% (95% CI 11.0–26.4%). GEO and AEO platform queries still cite off-site surfaces heavily.
Read the articleReddit led citations at 33.7% (95% CI 24.9–43.8%). Same-day variance flipped 10 of 10 panel queries.
Read the articleA citation rate here is the share of live engine answers that named a given domain. We report Wilson 95% confidence intervals on those shares. A variance panel repeats the same queries the same day so flip rates sit next to the headline.
GEO (Generative Engine Optimization) and Answer Engine Optimization are the category labels under study. Engines covered are ChatGPT, Gemini, and Perplexity. These sweeps do not cover Google AI Overviews.
Start with a free Fit Check — we measure where you stand, then recommend the package that fits.
Join waiting listCookies & analytics
Essential cookies keep you signed in. Analytics load only with your OK. Privacy Policy
