How to Check Your AI Visibility

"AI visibility" sounds abstract until you break it into things you can actually inspect. There is no dashboard equivalent to a Google Search Console ranking report — no single number a platform hands you. But the underlying causes of whether an AI assistant recommends a business are concrete, checkable, and mostly outside the business's marketing department: they live in robots.txt, in schema markup, in review platforms, in page copy, and in directory listings.

This is the same five-signal method AEO Radar's free scan runs automatically. Here's how to check each signal by hand, what a pass or fail actually looks like, and which signal is worth fixing first.

The five signals, in order of impact

Signal 1 — Crawler access (the gate)

Before anything else matters, the AI has to be able to read the site at all. Check your robots.txt file (yourdomain.com/robots.txt) for explicit rules covering GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, and Bytespider — the crawlers behind ChatGPT, Claude, Perplexity, Google's AI features, Common Crawl (a training-data source multiple providers draw from), and ByteDance's AI products, respectively. A blanket Disallow: / with no carve-out for these bots means the AI literally cannot see your pages. This isn't a ranking factor to optimise — it's binary, and a large share of businesses fail it without realising, because default WordPress/Squarespace/Wix robots.txt configurations were written years before AI crawlers existed and never got updated.

Signal 2 — Structured schema (the clarity layer)

LocalBusiness or Organization JSON-LD markup tells an AI system unambiguously what your business is, where it operates, and when it's open — rather than requiring the model to infer this from unstructured prose, which it can get wrong. Check by viewing your page source and searching for application/ld+json, or by running your homepage through Google's Rich Results Test. A controlled Ahrefs study (1,885 pages, difference-in-differences methodology) found schema markup's independent effect on AI citation rate ranged from −4.6% to +2.2% — a range that overlaps zero.¹ Schema is worth having for accuracy and rich-snippet eligibility. It is not, on its own, a citation driver — don't over-invest here relative to the signals below.

Signal 3 — Reviews and UGC (the strongest measured signal)

Among everything studied, review data correlates most strongly with AI recommendation. SOCi's 2026 Local Visibility Index (350,000+ locations) identified approximate rating floors below which each platform rarely recommends a business: ChatGPT around 4.3 stars, Perplexity around 4.1 stars, Gemini around 3.9 stars.² Recency compounds rating: BrightLocal's 2026 survey found 74% of consumers only trust reviews from the last three months, and 32% only trust reviews from the last two weeks.³ Check your current rating on Google and your primary directories, and check the date of your most recent review — a 4.6-star average with a review from eight months ago is a weaker signal than a 4.3-star average with reviews from this month.

Signal 4 — Content structure (the extraction layer)

AI systems quote and cite content that directly answers a question in an extractable format — a question as a heading, followed immediately by a direct, specific answer. A Princeton GEO (Generative Engine Optimisation) study found content containing citations, statistics, and direct quotes earns approximately 40% more AI visibility than equivalent content without cited evidence.⁴ Check your own site's key pages: do they lead with vague, adjective-heavy marketing copy ("award-winning service in a welcoming environment") or with specific, checkable claims ("Invisalign Diamond Provider since 2019, 340 verified Google reviews")? The second format is what gets extracted and quoted.

Signal 5 — Directory presence (the sourcing layer)

Different AI platforms source local citations from different places. Yext's citation-sourcing analysis (6.8 million citations) found ChatGPT draws roughly 49% of its local citations from third-party directories, Yelp chief among them, while Gemini draws roughly 52% from brand-owned websites.⁵ Check whether your business has claimed, accurate profiles on Yelp, Bing Places, and Facebook — an unclaimed or stale listing is worse than no listing, because it can surface outdated information as if it were current.

What to do with the result

Once you've checked all five, the pattern usually points at one or two clear gaps rather than a uniformly weak or strong profile. A business can have excellent schema and zero AI visibility because its Yelp profile is unclaimed. Another can have a 4.7-star rating and still be invisible to ChatGPT specifically because its content never answers a question directly enough to quote. The fix is almost always narrower than "improve everything" — it's usually one or two of the five signals doing most of the damage.