BLOG
Showing Analysis posts · 49 articles
Can you trust AI prompt volume data? What OpenAI has actually disclosed, how third-party estimates are built, where two studies disagree by 53 points, what vendors say about their own numbers and how to read any volume claim.
Does schema markup raise AI citations? Google's documentation, Ahrefs' 1,885-page experiment and an independent preprint compared, with sample sizes, dates, where they disagree and what to do about it.
Does ChatGPT recommend brands from memory or from live search? What Ahrefs, Semrush, OpenAI and independent tests show about how often it searches, why the numbers disagree, and what it means for brands.
Google click-through rate by position in 2026 from Advanced Web Ranking, First Page Sage, Ahrefs, Seer, Backlinko and Pew: how far CTR has fallen with AI Overviews, and why the studies disagree.
Do AI recommendations lift branded Google search? What Similarweb, Ahrefs and Semrush data show, where the studies disagree, why it's hard to measure, and how to check your own numbers.
Record funding, an Adobe acquisition and prices from free to $4,500 a month, while the answers these tools measure change from one run to the next. Where the AEO tools market stands, and what to check before you buy.
Which AI engine sends the most visits to websites, from SE Ranking, Statcounter, Conductor and Similarweb data, plus our own country-by-country check of 15 markets.
Where in an article AI engines pull citations from, from the CXL, Ahrefs, Previsible and Princeton studies: the top third, self-contained blocks, headings, length, and where the data disagrees.
Does traffic from ChatGPT and other AI assistants convert better than search? We lined up eight studies, from a one-site 23x to a 973-store paper where organic search won, and checked each at the source on October 19, 2026. The multiple shrinks as the sample grows, and the 2026 data looks better than 2025's.
Which page types AI engines cite most, from three large studies and our own October 2026 sample of 744 citations across 135 questions in ChatGPT and Google AI Mode. Listicles lead overall, but the question decides the winner.
How often ChatGPT searches the web depends on the question and the mode. We pooled 117 ChatGPT answers from October 2026 tests and read the public studies, including one that found Thinking mode ran nearly 5x more web searches than Instant.
AI hallucination rates in 2026 run from under 2% to over 70%, depending on the test. We pulled the October 2026 leaderboards and studies, then tested ChatGPT and Google AI Mode on 26 companies, 8 SaaS prices and 9 brands that don't exist.
THE SIGNAL
Get the latest on AI citations, Google rankings, and content strategy.
No spam. Unsubscribe anytime.