Your Brand Scores 31 Out of 100 in AI Recommendation Tests — Here Is How to Measure and Fix It
Independent studies show the average brand scores 31 out of 100 in AI recommendation tests, and citation rates swing by 24 points between engines. Here is the operator playbook for measuring and fixing your AI recommendation score.
The average brand scores 31 out of 100 when tested across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. That number comes from running 50 real buyer prompts across five AI engines and counting how often each brand appears in the answers. If you are not measuring this, you are managing your AI visibility blind.
The 31-out-of-100 problem
Most brands cluster between 20 and 40 on AI recommendation scores. That means your company appears in roughly a third of the prompts where a buyer could find you — not zero, but nowhere close to "being recommended." The Inithouse study found that brands scoring 80 and above consistently share three traits: structured data their sites expose for AI parsing, third-party mentions across multiple authoritative sources, and a clear positioning statement the engine can extract and repeat.
That last point matters more than most teams realize. If your product description reads like every competitor's, no AI engine has a reason to pick yours from the pile.
Why single-engine tracking gives you a false map
A Visiby benchmark of 172 buyer prompts across three engines found the median brand was cited in 11 percent of ChatGPT answers, 9 percent of Perplexity answers, and just 4 percent of Google AI Overviews answers. The same brand's citation rate swung by up to 24 percentage points depending on which engine answered the question.
The numbers get worse when you look at source overlap. On 11 percent of prompts where all three engines returned cited sources, they shared no source at all — three answers to one question, built on entirely separate domain sets. Reddit appeared in 45 percent of ChatGPT answers and zero Perplexity answers. YouTube showed up in 66 percent of Perplexity answers and less than 1 percent of ChatGPT's. Optimizing for one engine and assuming it transfers is a mistake I see teams make constantly.
Brand strength gets you recognized, not recommended
A FrictionAI study of 14,140 AI answers across five LLMs confirmed something I have suspected for a while: Knowledge Graph strength predicts whether an AI engine recognizes your brand, but it does not predict whether the engine recommends you. Recognition and recommendation are two different outcomes, and most brands only measure the first one.
This matters for anyone running an AI visibility strategy. You can be present in the training data, show up when someone asks "what is [your company]," and still be absent from every buying prompt in your category. The gap between being known and being chosen is where revenue lives.
The three-move diagnostic
Based on the converging data from these studies, here is the audit I would run this week:
1. Measure per-engine citation rate. Run your top 10 buyer prompts through ChatGPT, Perplexity, Gemini, and Google AI Overviews. Count how often your brand appears in each engine's answers separately. A single average hides the engines where you are invisible.
2. Check recommendation versus recognition. Ask each engine "what is [your company]" and then "recommend a [your category] tool for [your ICP]." If you appear in the first answer but not the second, you have a recommendation gap. Fix it with third-party mentions on sources each engine trusts — the Visiby benchmark shows those source preferences differ substantially between engines.
3. Audit your positioning clarity. If your homepage and product pages use the same language as every competitor in your space, AI engines cannot differentiate you. The brands scoring 80+ in the Inithouse data had distinctive positioning statements the engines could extract and repeat verbatim.
What this means for your AI visibility investment
The measurement infrastructure for Machine Relations — the relationship between your brand and the AI engines that mediate buyer discovery — is still early. But the data is clear enough to act on. If you are spending time and budget on AI visibility without measuring recommendation rates per engine, you are investing without a feedback loop.
The delta between your best and worst engine score is probably 30 points or more. That delta tells you exactly where to focus: not on the engine where you already show up, but on the one where buyers ask and you are absent.
FAQ
How do I measure my brand's AI recommendation score?
Run 10 to 50 real buyer prompts — the kind your customers actually type — through ChatGPT, Perplexity, Gemini, and Google AI Overviews. Count how often your brand appears in each engine's response. Your recommendation score is the percentage of relevant prompts where your brand shows up, measured per engine. Tools like Be Recommended and Visiby automate this process across multiple engines.
Why does my brand appear in ChatGPT but not Perplexity?
Each AI engine uses different source preferences and retrieval architectures. Visiby's benchmark found that Perplexity's source universe is roughly 40 percent larger than ChatGPT's, and the two engines share sources on fewer than 12 percent of prompts. A brand can rank well in one engine's preferred sources while being absent from another's. Fixing this requires building authority on the specific third-party sources each engine trusts.