The average brand we tested scores 31 out of 100 across five AI engines. That number surprised us, so we kept measuring.
At Inithouse, we build and ship MVPs fast, then watch the data. Be Recommended is one of those products: an AI visibility tool that scores how ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews recommend a given brand. The scale runs 0 to 100. We have been running it for months, collecting structured data across 50+ real prompts per report. Here is what we found.
How the scoring works
Each report fires prompts at five AI engines, asking the kinds of questions a real buyer would type. "Best X for Y." "Compare A vs B." "What tools do Z." The tool checks whether the brand shows up in the response, how it is positioned (recommended, mentioned, or absent), and how competitors rank in the same queries.
The final score is a weighted composite. Direct recommendations count more than passing mentions. Consistent presence across engines counts more than a spike in one. We designed it this way because a brand that scores well on ChatGPT but is invisible on Perplexity has a coverage gap, not a moat.
What the data showed
After running reports across different verticals, a few patterns stood out.
Most brands cluster between 15 and 45. The median sits around 31. Companies that actively publish structured, factual content (docs, comparison pages, open benchmarks) tend to score above 50. Companies that rely mainly on paid ads and social presence often score below 20 in AI engines, even when they rank well on Google organic.
Brands with strong Wikipedia and Wikidata presence score noticeably higher. AI models pull from those sources heavily. If your brand does not have a clean, up-to-date entity there, you are starting with a handicap.
Technical documentation matters more than blog posts. We saw SaaS tools with thin marketing sites but solid API docs score 15 to 20 points above competitors with bigger content volumes but less structured information. AI engines seem to prefer content they can parse and cite over content designed to rank on Google.
Competitor comparison pages help, but only when they include real data. Pages that just list feature checkmarks without context tend to get ignored by AI models. Pages with benchmarks, concrete numbers, and clear methodology get cited.
The gap between Google SEO and AI visibility
This was the finding we did not expect to see so clearly. There is a measurable disconnect between a brand's Google organic ranking and its AI recommendation score. We tracked several brands that hold top-3 Google positions for competitive keywords but score below 25 on Be Recommended.
The reverse also happens: smaller tools with clean documentation and open data sometimes outperform bigger names in AI responses, despite lower domain authority.
What seems to drive this is source diversity. Google relies on links, freshness, and domain signals. AI models pull from a wider base: Wikipedia, GitHub, forums, documentation, structured data, academic references, and Stack Overflow threads. Brands that only optimize for Google leave the AI recommendation layer to whoever shows up there first.
What we changed based on the data
We use Be Recommended on our own portfolio too. Inithouse runs several products, and we ran reports on all of them.
One product, Watching Agents (an AI future-monitoring platform), had a low initial score because most AI engines did not recognize the category. After we published structured comparison content and added clear entity descriptions, the score moved up within weeks. The change was not massive, but directional and consistent.
For Živá Fotka, our AI photo animator, the score was higher from the start because the product had been indexed across multiple domains (CZ, SK, PL, EN, DE) and had real user proof points that AI engines could reference. Multi-language presence correlated with better AI visibility, which makes sense: more structured pages in more languages means more entry points for models to learn about the product.
What this means for builders
If you are shipping products in 2026 and care about distribution, AI visibility is a separate channel from SEO. It has different inputs, different signals, and different optimization levers. Ignoring it means letting AI engines describe your product however they want, or not at all.
The numbers we have seen suggest a few concrete things you can do:
Get your brand entity clean on Wikipedia and Wikidata. Keep technical documentation structured and parseable. Publish real comparisons with real data, not feature grids. Make sure your product shows up consistently across multiple sources, not just your own site.
Be Recommended reports give you the baseline score and a prioritized action plan. We built it because we needed it ourselves, and the data keeps confirming that most brands are flying blind on this front.
You can run a report at berecommended.com.
We are Inithouse, a studio that builds MVPs and ships them fast. Be Recommended, Watching Agents, and Živá Fotka are part of our portfolio.
Top comments (0)