The average brand scores 31 out of 100 when we measure how often AI engines recommend it. That number comes from running 50+ real user prompts (the kind people actually type) across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews, then counting how many times a brand shows up in the answers.
We built Be Recommended at Inithouse to do exactly this measurement. The tool scores AI visibility on a 0-to-100 scale across all five engines and produces a prioritized action plan for getting recommended more. Here's what the data taught us so far.
What "citation rate" actually means
When someone asks ChatGPT "what's the best project management tool?" or "recommend a good CRM for small teams," the AI pulls from its training data, web results, and whatever retrieval-augmented pipeline it uses. Some brands appear in that answer. Most don't.
Citation rate is the percentage of relevant prompts where a given brand appears in the AI's response. A brand with a citation rate of 80% shows up in 4 out of 5 prompts where it could reasonably be mentioned. A brand at 10% appears in 1 out of 10.
The gap between those two numbers is the gap between being recommended and being invisible.
The method: 50+ prompts, 5 engines, real language
We don't use synthetic or academic prompts. Each prompt set mirrors actual user queries: buying-intent, comparison, "best X for Y," problem-solving. The kind of thing people type into ChatGPT instead of Google these days.
Every prompt runs against five engines:
- ChatGPT (GPT-4o)
- Claude (Anthropic)
- Perplexity (with citations)
- Gemini (Google)
- Google AI Overviews (Search Generative Experience)
The responses get parsed for brand mentions, and each brand receives a composite score weighted across engines.
Three findings that surprised us
1. Most brands cluster between 20 and 40.
The average across all reports we've generated sits around 31. That means most companies get mentioned in roughly a third of the prompts where they could appear. Not zero, but not enough to call it "being recommended."
2. Engine variance is massive.
A brand might score 65 on Perplexity (which cites sources aggressively) and 12 on Gemini (which tends to favor Google's own ecosystem in certain categories). Optimizing for one engine doesn't mean you're visible on another. The delta between a brand's best and worst engine score averages 30+ points.
This was the finding that shaped how we built the tool. A single "AI visibility score" without per-engine breakdown would hide the most actionable signal. You need to know where you're invisible, not just that you are.
3. The top brands share three traits.
Brands scoring 80+ consistently have: (a) structured data on their site that AI can parse, (b) third-party mentions across multiple authoritative sources, and (c) a clear, unique positioning statement that AI can parrot back. If your product description reads like every competitor's, the AI has no reason to pick yours.
What this means for builders
If you're shipping a product in 2026, your SEO checklist needs an AI visibility line item. Traditional search rankings still matter, but a growing chunk of discovery is happening through AI chat interfaces where the rules are different.
At Inithouse, we run AI visibility checks across our own portfolio to understand which products get mentioned and which don't. Be Recommended came out of that internal need. We wanted a number, not a vibe.
The actionable takeaway: measure your AI citation rate, identify which engines miss you, and fix the inputs those engines rely on. That means your site structure, your third-party footprint, and your positioning clarity. The brands doing this now are building a lead that compounds.
Try it
Be Recommended generates a full AI Visibility Report: your score across all five engines, a competitor comparison, and a prioritized action plan. We built it because we needed it ourselves.
If you want to see where your brand stands, run a report and compare your numbers to the 31-point average. The gap between "mentioned sometimes" and "recommended by default" is measurable now, and that makes it fixable.
Top comments (0)