Knowledge base›Sources›Josh Blyskal (Profound)
Blyskal analyses 250M answers and finds: most prompts trigger two to three search queries; classic SEO signals explain almost nothing of citation probability; overlap with Google is 39%; half the cited pages are younger than a quarter; on product pages FAQs raise the citation rate by 848%. In the follow-up (July 2026) he adds that ChatGPT appends year numbers in 17% of fan-outs and reaches 37% alignment with Google's top 50.
Profound is the largest output statistic in the field. Their numbers are the yardstick our mechanics findings can be measured against, and their dated breaks (entity update 2025-10-18, 5.3 instant 2026-03-04) are exactly the places where a tracker with schema diffs could have named the cause.
His fan-out figures are half a year old and below ours; that matches Peec's observation of growing fan-outs. The freshness statement our windows explain mechanically: a window of one to 30 days lets no old page through. The Google overlap is not measurable as a number for us, but the pipeline signature shows for which topics Google is the source. What Profound does not see, we do: the mechanics behind the numbers.
| Claim | Status | Evidence |
|---|---|---|
| 36.4% of prompts trigger two, 52.9% three search queries | outdated | In our data 2-7 per searching question, median 3 (2026-08-22). Trend higher; Peec also measures rising fan-out length since Jan 2026. |
| Classic SEO metrics explain 4-7% of citation variance | unverifiable | No brand monitoring in the project. |
| 39% overlap with Google top results | unverifiable | No SERP comparison. But our pipeline signature shows that news and price questions are served directly from Google scrapes - there the overlap is high by construction. |
| 50% of the most-cited pages younger than 13 weeks | confirmed | Freshness windows in the fan-outs (1-90 days) force recency for news, prices, products. For evergreen topics no search happens at all. |