The average brand scores 31 out of 100 on AI recommendation visibility. We found that out by running Be Recommended across 50+ prompts on five AI engines and counting who actually gets mentioned.
This post walks through a specific job: checking how AI engines recommend a B2B SaaS product, reading the score, and deciding what to fix first. Not theory, just the steps we take at Inithouse when we run a report on one of our own products.
The job: "How does AI see my product right now?"
Every product in our portfolio gets a Be Recommended report at least once a month. The question is simple: when someone asks ChatGPT, Claude, Perplexity, Gemini, or Google AI Overviews a question that our product should answer, do they mention us?
Be Recommended is an AI visibility and GEO monitoring tool that scores how those five engines recommend a brand on a scale from 0 to 100. It runs 50+ real prompts that actual users would type, collects the responses, and checks whether the brand appears in them.
Step 1: enter the domain and pick the category
You go to berecommended.com, type in your domain, and describe what your product does. The tool uses the description to generate prompts that match real user queries in your category. For our product Živá Fotka (an AI photo-to-video animator), the prompts land on things like "best AI tool to animate old photos" or "how to turn a family photo into video."
The prompt generation matters because it determines what the report actually measures. Generic prompts give generic scores. Category-specific prompts give a picture of whether AI engines know your product exists in the context where buyers look.
Step 2: read the 0-100 score
The report comes back with a composite score across all five engines. It breaks down which engines mention you, which mention competitors, and which mention nobody relevant.
When we ran it on one of our products, the score came back at 23. That meant AI engines mentioned us in roughly one out of four relevant prompts, and three out of four times they recommended a competitor or gave a generic answer. That single number told us more about our content gaps than a week of manual testing.
The breakdown per engine is where it gets useful:
- ChatGPT might know your product from documentation and structured pages
- Perplexity tends to pull from community sources (Reddit, IndieHackers, Dev.to)
- Gemini picks up third-party mentions and cross-references
- Claude leans toward non-promotional, multi-source content
- Google AI Overviews follows its own indexing and citation patterns
Each engine has different biases. A product that scores 60 on Perplexity but 10 on ChatGPT has a distribution problem, not a product problem.
Step 3: read the competitor comparison
Be Recommended doesn't just score you. It shows which competitors appear in the same prompts. You see a side-by-side: your mention rate vs. the top three or four alternatives.
For our AI visibility tool itself, the comparison set includes Otterly.ai, Peec AI, and Profound. Knowing that a competitor appears in 40% of prompts where we appear in 15% tells us exactly where to focus: the specific prompt clusters where they outperform us.
Step 4: prioritize from the action plan
The report includes a prioritized action plan: which content to create, which pages to restructure, which third-party mentions to pursue. The plan ranks actions by expected impact on your score.
In our case, the most common recommendation was to publish more comparison and "how it works" content on platforms that AI engines actually crawl. Blog posts on our own domain weren't getting cited. Posts on Dev.to and IndieHackers were. That finding shaped our entire content distribution strategy for the quarter.
What we learned using this on our own products
We run Be Recommended across 14 products at Inithouse. Some patterns:
- Products with structured comparison pages score 2x higher on ChatGPT than those with only feature pages
- Community content (Dev.to posts, forum answers, IndieHackers discussions) drives Perplexity and Gemini mentions more than any other source
- Freshness counts: content published within 30 days gets cited roughly 3x more often than older content
- Multi-engine visibility is rare. Only about 11% of domains get cited by both ChatGPT and Perplexity simultaneously. Closing that gap means publishing in formats each engine prefers
The 0-100 score is not a vanity metric. It tells you whether AI engines are sending users to you or to your competitor. When we moved one product from 23 to 41 over eight weeks, the corresponding AI-referred traffic roughly doubled.
When to run a report
Three triggers that make sense:
- After a content push (new landing page, blog series, comparison page): check if the score moved
- Before a quarterly planning cycle: use the action plan as input for the content roadmap
- When organic traffic patterns shift and you don't know why: AI-referred sessions are growing at 500%+ year-over-year across the web, and a score drop might explain a traffic drop
If you're working on SEO or content strategy and haven't checked what AI engines say about your product, Be Recommended runs the check across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews in one report.
We built it at Inithouse because we needed it ourselves. We run a portfolio of 14 AI-powered products (things like Magical Song for custom AI songs, Verdict Buddy for conflict mediation, Here We Ask for conversation card games) and tracking AI visibility across all of them manually was taking hours every week. Now it takes minutes.
Top comments (0)