How do you audit a website for AI search readiness?

A website can pass every classic technical SEO check and still be functionally invisible to ChatGPT, Google's AI Overviews and Perplexity. The checks aren't the same, because these tools retrieve, parse and cite content differently to how a classic search algorithm ranks it. Most existing audit templates simply weren't built with this in mind, which means a genuinely useful AI readiness audit needs a few additions most standard checklists still miss.

None of these additions require throwing out an existing technical SEO process. They layer on top of it, which means the most efficient path for most businesses is extending an audit they already run, rather than commissioning something entirely separate and duplicating work that's already being done well.

Step one: confirm AI crawlers can actually access the site

This step takes minutes and should always come first, since every other finding in the audit is meaningless if the content being evaluated can't actually be retrieved by the tools in question.

Check robots.txt directly for directives blocking GPTBot, Google-Extended, PerplexityBot, ClaudeBot and similar crawlers. This is the single most common and most easily overlooked issue, since these directives are sometimes added deliberately during a security review, or left over from a staging environment configuration that never got reverted after launch. A site can rank perfectly well in classic search while being entirely blocked from AI retrieval because of one line nobody checked.

Step two: check how the content is actually structured

This step often reveals the biggest, most immediately actionable gap, since restructuring an existing page to lead with a clear answer is usually far quicker than producing entirely new content from scratch.

AI models favour content that answers a specific question directly and clearly, ideally near the top of the page, over content that builds up to a point across several paragraphs. Audit key pages by asking: does this page answer one clear question in the first few sentences, or does a reader, or a model, need to read the whole thing to extract the actual answer. Pages written as long, meandering narratives are far less likely to be lifted and quoted than pages structured around clear, direct answers.

Step three: review structured data specifically for AI parsing

FAQ schema, article schema, organisation schema and author schema all give AI models explicit, structured signals about what a page contains and who's behind it, on top of whatever benefit they provide for classic search results. An audit should check not just whether schema exists, but whether it's implemented correctly and validates without errors, since broken schema often provides no benefit at all.

Step four: check entity consistency across the web

AI models build confidence in a business based on how consistently it's described across multiple sources, not just its own website. An audit here means checking whether your business name, services and key facts are described consistently across your site, your Google Business Profile, industry directories and any press coverage, or whether there are contradictions and inconsistencies that could make a model hedge rather than confidently recommend you.

1. 01Search your business name plus your core service and see what AI tools currently say about you
2. 02Compare how your business is described across your website, Google Business Profile and any directories
3. 03Note any factual inconsistencies: different service descriptions, outdated information, or conflicting details
4. 04Check whether competitors are being named and cited instead, and note which sources those citations come from

Step five: assess unlinked mentions and third-party coverage

Traditional SEO audits focus heavily on backlinks specifically. An AI readiness audit needs to look more broadly at any mention of the business, including ones without a hyperlink, since language models weigh unlinked mentions on trusted sites as evidence too. A business named favourably in a trade publication without a link still benefits from that mention in a way classic link-based SEO metrics wouldn't capture at all.

Step six: test actual retrieval, not just theoretical readiness

The most direct test is simply asking ChatGPT, Copilot and Perplexity the real questions a buyer would ask, and recording whether your business appears, what's said about it, and which sources get cited. This turns an audit from a checklist of theoretical best practices into a genuine, current measurement of where things actually stand.

What this looks like in practice

A B2B software company runs through these steps and finds robots.txt is clear, a genuinely good sign, but its content is written in long, scene-setting introductions before ever answering the actual question in the page title. Its schema markup exists but throws validation errors on half the site due to a template update eighteen months ago that nobody noticed broke it. And when tested directly in step six, ChatGPT names two competitors by name for a core service question and doesn't mention the company at all. None of these findings alone would explain weak AI visibility. Together, they show exactly where the effort needs to go: fix the schema errors first since they're quick technical wins, then restructure key pages to answer questions directly, then work on the entity and citation gap the direct testing revealed.

Common objections, answered

We already ran a technical SEO audit, isn't AI readiness covered by that?

Partially. A standard technical audit typically covers crawlability, site speed and structured data at a general level, all relevant here too, but it rarely checks specifically for AI crawler directives, answer-shaped content structure, or entity consistency across external sources, which are the additions that actually determine AI-specific visibility.

How do we know if a drop in AI mentions is even worth worrying about?

If AI-driven research plays any meaningful role in how your buyers choose a supplier, and increasingly it does across most B2B categories, then yes, it's worth tracking and addressing. The businesses that treat this as unimportant now are the ones most likely to be caught off guard once it becomes an obvious, visible gap in their pipeline.

Can small businesses realistically compete for AI citations against larger competitors?

Often more easily than they can compete for broad organic rankings. AI models favour clear, specific, well-documented expertise over sheer domain size, which gives a smaller, genuinely specialist business a real chance to be cited over a larger, more generic competitor for the specific questions its actual customers ask.

Turning the audit into action

The findings from each of these steps point to different fixes: technical changes for crawler access and schema, content restructuring for answer-shaped pages, and outreach or directory updates for entity consistency. Prioritising these by how directly they affect actual citations, tested through the real questions in step six, is what turns an audit into a genuine improvement plan rather than a static report. An AI SEO programme should be built around exactly this kind of audit as its starting point, not a generic technical review with AI mentioned as an afterthought.