Measurement essay12 min
How I measure AI visibility
Why I read visibility rank before score, how citations reveal what shaped an answer, and where attribution breaks.
Josh Blyskal
I make brands visible in AI search.
I joined Profound as its second employee and now lead AI Strategy & Research there. I compare how ChatGPT, Claude, Gemini, and Google AI products research products, resolve brand identity, and decide which brands appear in an answer.
I use research drawn from Profound's proprietary datasets—more than 10 billion AI citations and 1.5 billion real user prompts—to map the questions, product facts, and public pages that determine whether a brand appears. The published findings include the samples and limitations I can share publicly.
I study what makes a brand appear in an AI-generated answer.
That includes the prompts people ask, the fanouts an agent/answer engine runs, and the websites, product data, and public information it uses to understand and recommend a brand to a user.
That work grew into the methodology Profound uses with enterprise teams. What I learn ends up here, with the sample and limitations attached.
Browse all research →Original research
Research11 min
The state of AEO in 2026: Claude is not ChatGPT
Across roughly 35,000 URLs associated with 400 queries, 79.2% of Claude's cited URLs also appeared in Brave's top 10.
Research12 min
What 250 million AI search results say gets cited
Traditional SEO metrics explained only 4% to 7% of citation variance across 1,311 pages.
SAGE for AEO
I built SAGE at Profound to keep AEO work in order: establish the baseline, diagnose the constraint, ship the response, then make the elements that had the highest impact repeatable.
Read the full method →- 01Setup
- 02Analyze
- 03Generate
- 04Engineer
Imagine the team tracks 20 prompts every week and the brand suddenly stops appearing in ChatGPT for five of them. Instead of immediately writing a new page, open those five responses and see what changed: which competitors now appear, which websites ChatGPT used, or maybe whether it searched for a different fanout phrase. SAGE exists to keep that diagnosis ahead of enterprise content calendars.
Selected work and press
We analysed 10,000,000 AI search results: here's what we found
A BrightonSEO keynote on why the same brand can appear differently across ChatGPT, Google AI Overviews, Perplexity, and Copilot, and what marketers can do about it.
Watch the full talk ↗
brightonSEO · Brighton, United Kingdom · April 2025
If you want to talk, I'm easy to reach.
For conference bookings, research collaboration, or commentary on AI search, email me directly.
josh@tryprofound.com