About this research

One researcher, a fixed method, and the same questions asked the same way every time.

AI Visibility Gap is run by a single independent researcher, not an agency or a research team. Every study follows the same buyer conversation format, tested manually across the same three AI platforms, published with the full dataset attached.

Who runs this

David Wood is an independent researcher based in Scotland, working on how AI systems form commercial recommendations. Before AI Visibility Gap, he ran the 2026 AI Citation Visibility Study for Crypto Protocols, testing 50 protocols across ChatGPT, Perplexity, and Google AI Overviews, and produced the Generative Engine Optimisation consultancy work at CryptoContent.dev.

That earlier study is where the method for this site came from. One detail from it is worth mentioning directly: raw Perplexity citation counts included dozens of Cloudflare CAPTCHA pages and privacy-footer links picked up by automated scraping, not real citations at all. They had to be manually removed before the numbers meant anything. It's the kind of error that only shows up once you sit and read every citation yourself, and it's why every study on this site is checked by hand rather than pulled automatically and trusted.

How each study is built

Every study runs the same three-stage process. Nothing is scored by hand and nothing is inferred without a recorded observation behind it.

1
Observation
A set of realistic buyer conversations, written as a real prospect would ask them, run verbatim across ChatGPT, Perplexity, and Google AI Overviews. Each response is captured with the platform, prompt, date, model version, and raw answer text, so results can be reviewed and compared later rather than trusted from memory.
2
Analysis
Recommendations, citations, and the point in each conversation where AI starts naming suppliers are extracted from the raw observations and compared across companies in the same market.
3
Application
Patterns become a published benchmark: which companies get recommended, which sources AI is actually citing, and what a company would need to demonstrate to close the gap.

Credentials and prior work

For anyone checking who's behind this before linking to it.

ORCIDorcid.org/0009-0007-3353-7188
Study DOI10.5281/zenodo.20146677
Working paperInternet Archive
Academia.eduindependent.academia.edu/DavidWood342
SlideShareslideshare.net/DavidWood774240
LinkedInDavid Wood
X / Twitter@TheDavidWood

Get in touch

For research enquiries, data requests, or corrections: admin@aivisibilitygap.com. For consultancy work applying these findings to a specific company, that sits with CryptoContent.dev.