AI Visibility Audit

Access: can the agents actually fetch your pages

We start at the bottom of the stack because that is where silent failures live. Robots.txt is read per agent — GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Perplexity-User, Bingbot, Googlebot — and the Google-Extended token is checked separately since it governs Gemini use rather than crawling. Then we look past the config file at reality: CDN bot rules, WAF challenges, rate limits, geo blocks and Cloudflare settings that return 403s to agents your robots.txt technically welcomes. Server logs settle the argument. We report which agents visited in the log window, which were refused, and what status codes they received.

Index coverage: are you even a candidate

An answer engine can only cite what its retrieval layer can find. We check Google indexation through Search Console coverage, and Bing indexation through Bing Webmaster Tools, which matters because Copilot grounds on Bing's index and ChatGPT has long drawn on Bing results alongside its own crawling. Missing sitemaps, unverified Bing property, noindex left on after a migration, canonical tags pointing at the homepage and orphan pages with no internal links all surface here. This section is frequently the whole answer for sites that assume they have an "AI problem" when they have an ordinary indexation problem.

Identity: one entity or three

We map how your business and your people are described across the site and off it. That means the Organization node and whether it uses one stable @id or a new one per page, the Person node for authors and whether a real profile page backs it, sameAs links and whether the profiles they point to actually exist, and NAP consistency across the site, the Google Business Profile and any directory listings. Fragmented identity is the most common finding in Indian SMB sites: a name written four ways, a phone number changed on one page, an author byline with no bio anywhere. Models cannot attribute claims confidently to something they cannot pin down.

Extractability: can a model quote you cleanly

We score your top pages on the properties that decide whether a passage can be lifted: does an answer appear in the first hundred words, is the subject named rather than implied, are figures dated and unit-bearing, does each page hold to one topic, are there tables where a comparison is the real question, and is critical content present in the HTML rather than injected by JavaScript. Alongside this we validate structured data — Article, Organization, Person, FAQPage, Breadcrumb — and flag markup that is present but wrong, which is more common than markup that is missing.

Baseline, deliverable and what it will not tell you

We agree a prompt set of twenty to forty real buyer questions and record, verbatim, how ChatGPT, Gemini, Perplexity and Copilot answer them today, including which domains get cited. That is your baseline; without it every later claim of improvement is unfalsifiable. The deliverable is a prioritised fix list with owners and effort, not a 90-page PDF. Price is ₹10,000 one-time, delivered remotely anywhere in India. What it will not do is predict citations or produce a "visibility score" pretending to be an industry benchmark — no such benchmark exists, and inventing one would be dishonest.

AI Visibility Audit — FAQs