A free 9-step AI visibility audit any brand can run

You can run a real AI search visibility audit without paying for software. A browser, a spreadsheet, and a few disciplined prompts are enough to reveal the big failure points: AI systems cannot reach your pages, cannot clearly identify your brand, cannot extract usable facts, or cannot trust what they find. That is already enough to start a serious Generative Engine Optimization (GEO) process.

What a free audit will not give you is trend data, prompt history, or cross-model benchmarking at scale. But for a first diagnosis, it is more useful than many teams think, especially if your current assumption is still, "We rank on Google, so AI must know us too."

• AI visibility is not the same thing as Google ranking.
• A strong free audit checks access, entities, content specificity, proof, and consistency.
• The highest-leverage fixes are often basic: blocked crawlers, vague copy, and conflicting facts.
• Manual audits are great for diagnosis, but weak for repeatable tracking over time.

Why do traditional SEO audits miss AI visibility problems?

Because classic SEO audits mostly ask whether a page can rank. AI visibility asks whether a system can find, understand, trust, and reuse your brand in an answer. Those are related questions, but they are not the same question.

A site can have solid rankings, decent backlinks, and healthy Core Web Vitals, yet still disappear from ChatGPT, Claude, Perplexity, or Google AI experiences. The gap often sits in entity clarity, fact consistency, extractability, or proof. If your site sounds polished to humans but vague to machines, you can still lose.

This is not theoretical. According to SOCi's 2026 Local Visibility Index, ChatGPT recommended just 1.2% of the local businesses it studied, versus 35.9% visibility in Google's local 3-pack. In retail, SOCi found only about 45% overlap between the brands most visible in traditional local search and the brands AI platforms recommended. That is the key mindset shift: strong SEO can help, but it does not guarantee AI inclusion.

A useful way to think about it is this: AI visibility usually breaks on one of three layers. Retrieval fails because the page is hard to access or parse. Brand understanding fails because the entity is weak or inconsistent. Recommendation fails because the brand has little proof, weak differentiation, or conflicting signals across the web. If you want the measurement framework behind that shift, BotRank's guide on how to measure your AI visibility is a good next read.

What should a free AI visibility audit actually check?

It should check the same things an AI system implicitly checks before it mentions or cites you: can it access your content, resolve your identity, extract clear facts, verify your claims, and reconcile your information across sources. The best free audit is not a bag of tricks. It is a structured way to test whether your brand is machine-readable and recommendation-ready.

1. Can AI systems actually reach your site?

Start with access. Open your robots.txt file and look for rules that block AI crawlers such as GPTBot, ClaudeBot, PerplexityBot, or Google-Extended. Then test the homepage and a few key landing pages as if you were a first-time visitor with no cookies, no saved location, and no JavaScript assumptions.

The most common failures are boring but damaging: a cookie wall that covers the real content, a language or location selector that requires a click, or heavy client-side rendering that leaves little usable HTML on first load. A luxury hotel site with beautiful design but a mandatory country selector is a perfect example. Googlebot may handle it. Other systems may not.

If you want a simple hands-on method, BotRank recently explained how to test whether AI search can retrieve your page. It is a fast way to separate visibility problems from basic access problems.

2. Do you have a complete entity inventory?

AI needs more than a homepage headline. It needs a map of the actual things your business is made of: brand, products, services, founders, locations, policies, use cases, pricing logic, and important differentiators. This is where Entity SEO becomes practical rather than theoretical.

Most brands discover the same issue during this step: they mention important entities, but they do not define them clearly anywhere. A SaaS company may list integrations on a features page but never explain what each integration does, who it is for, or what problem it solves. A hotel may mention room categories without precise occupancy rules, accessibility details, or proximity facts.

Your audit should produce a simple entity sheet. List each entity, the page that defines it, the key facts attached to it, and the external profiles or references that reinforce it. If that list feels messy, that is already the diagnosis. BotRank's article on how to audit your AI entity footprint goes deeper on this exact exercise.

3. Is your content specific enough to quote?

AI systems quote and paraphrase specifics better than they interpret fluffy copy. If your page says, "We help teams move faster with a powerful platform," that sounds fine in a brand workshop and terrible in a retrieval pipeline. Faster than what? For which teams? Doing which job?

A simple test works well here. Paste a key page into an AI assistant and ask it to flag vague pronouns, empty adjectives, and sentences that do not stand alone. You are looking for phrases like "it," "this," "industry-leading," "seamless," or "best-in-class" with no concrete referent. Pages full of those phrases are hard to cite cleanly.

A stronger sentence looks like this: "Our onboarding software helps mid-market HR teams automate offer letters, e-signatures, and policy acknowledgments across six countries." That sentence gives the model something usable. It names the audience, the job, and the scope.

4. Can every important claim be traced to evidence?

Unsupported claims are a trust leak. If you say you are the best, the only, the fastest, or the most awarded, you should be able to show where that came from. If you cannot, the claim may help your homepage copywriter but it gives an AI system nothing stable to rely on.

This step is brutally simple. Highlight every superlative or proof-like claim on your most important pages. Then add the supporting source next to each one. That source might be a customer count, a named review platform, a certification body, an analyst report, an internal benchmark, or a published case study.

For example, "trusted by 1,200 finance teams" is stronger than "trusted by leading companies" if the number is current and can be backed up internally. Likewise, "SOC 2 Type II certified" is stronger than "enterprise-grade security" because it points to a concrete validation standard.

5. Are you saying anything competitors cannot say too?

This is the information gain test. AI systems do not need your page if it says exactly what ten other pages already say. If your copy is generic, your brand becomes replaceable inside the answer generation process.

Take three to five distinctive sentences from your site and search them in quotation marks. If similar phrasing could come from any company in your category, you have a differentiation problem. The cure is not cleverer adjectives. It is first-party specificity.

The best material here is usually proprietary and operational: a named methodology, a benchmark from your own customer base, a transparent comparison table, a specific implementation workflow, or a documented edge case. A consulting firm that explains its 4-week audit process in detail gives AI something reusable. A firm that says it delivers bespoke excellence does not.

6. Is your structured data helping machines understand the business?

Structured data should make the entity clearer, not just tick a compliance box. Right-click a page, view source, and search for application/ld+json. If nothing appears, you are asking machines to infer everything from prose. If something appears, the next question is whether it is thin, outdated, or disconnected from your actual site structure.

Look for basics first: does the markup identify the organization, the page type, products or services where relevant, and any authorship or FAQ structure that genuinely exists on the page? Then look for stronger entity signals such as a meaningful sameAs field and nested relationships that reflect how the business is actually organized.

You do not need to guess whether the syntax is valid. Run key URLs through Google's Rich Results Test. And if you want a recurring technical layer for this kind of review, BotRank's Technical Audits feature is built to track AI-readiness issues like crawler access, structured signals, and discoverability over time.

7. Do your core facts match everywhere online?

Consistency is one of the cheapest fixes in all of GEO. If your website says one thing, your Google Business Profile says another, your LinkedIn company page says a third, and a directory still shows an old phone number or old pricing model, AI systems have to reconcile the conflict. That usually lowers confidence.

Build a side-by-side sheet with your most important public sources. For local brands, start with hours, phone number, address, service area, and policies. For SaaS or B2B brands, include product description, target audience, headquarters, pricing entry point, integrations, and support availability.

A practical example: if your homepage says "starts at $49," your pricing page says "contact sales," and an old review site still lists a discontinued free plan, AI can easily synthesize the wrong answer. The issue is not that the model hallucinated. The web told three different stories.

8. Are you connected to recognized knowledge sources?

AI systems are better at selecting known entities than discovering ambiguous ones from scratch. That makes knowledge-source connections more important than many brands realize. Search your brand name and see whether a Knowledge Panel appears. Then check whether you have a traceable presence on sources like Wikidata or other authoritative profile surfaces relevant to your market.

This does not mean every company needs a glamorous encyclopedia footprint. It means the open web should help a machine answer basic identity questions with confidence: who you are, what category you belong to, where you operate, and what other trusted sources point back to you.

The practical audit question is simple: if your brand name is somewhat generic, what helps a system disambiguate you from others with a similar name? Sometimes the fix is as simple as stronger sameAs references, better organization markup, or more consistent use of the full brand name across profiles.

9. What do AI systems actually say about you today?

This is the moment where theory ends. Ask ChatGPT, Claude, Perplexity, and Google a small set of real buyer questions in a fresh browser session. Compare the answers against reality. You are not just looking for whether your brand appears. You are looking at accuracy, description quality, cited sources, and competitor presence.

Use prompt categories, not one-off vanity searches. Ask a branded question, a category question, a use-case question, and a comparison question. For example: "What does [brand] do?" "Best payroll software for multi-country startups." "How do I automate employee onboarding across regions?" "[brand] vs [competitor]." The mix matters because brands often perform well on branded prompts and disappear completely on non-branded ones.

Manual testing like this is the fastest way to establish a baseline. But once you need repeatability, you need a system that stores prompts and reruns them. That is exactly what BotRank's AI Visibility tracking is designed for.

Where does a manual audit stop being enough?

A manual audit stops being enough when the question changes from "Do we have a problem?" to "Is this improving, on which models, for which prompts, against which competitors, and because of which pages?" Diagnosis is manual-friendly. Ongoing measurement is not.

The biggest weakness of a free audit is variance. Ask the same prompt twice and you may get two different answers, two different cited sources, or two different competitors. A spreadsheet snapshot is still useful, but it is not a tracking system.

A good rule is this: if you are still discovering basic issues, stay manual for a week or two. If you are already making fixes and need to know whether those fixes changed visibility, graduate to tooling before your team starts arguing from anecdotes.

BotRank's Take: the checklist is useful, but the real bottleneck is repeatability

The smartest part of this 9-check framework is that it turns AI visibility into observable failure modes. That matters because many teams still treat GEO like a content volume game when the real issues are often crawler access, entity confusion, thin proof, or messy cross-source facts. A free audit forces clarity.

Where most teams hit the wall is not diagnosis. It is repeatability. Running a few prompts by hand can show that your brand is missing. It cannot reliably show whether last month's schema cleanup, new FAQ block, or rewritten product page actually changed visibility across models. That is where BotRank's Source Analysis becomes useful. It helps teams inspect which pages and external sources are actually supporting AI answers, so fixes stop being guesswork. In practice, that means you can distinguish "AI never reached our page" from "AI reached it but trusted another source more." That is a much better conversation than staring at one odd chatbot answer and calling it a trend.

How should teams turn this audit into an action plan?

Turn the checklist into a prioritization system. The goal is not to finish nine tasks for the sake of completeness. The goal is to fix the issues most likely to change recommendation outcomes on the pages and prompts that matter most to revenue.

• Score each check: use a simple 0, 1, or 2 scale for every important page or entity. Zero means missing, one means partial, two means strong.
• Start with high-intent pages: prioritize homepage, product pages, service pages, category pages, pricing, comparison pages, and top local landing pages.
• Group fixes by failure type: access issues, entity issues, proof issues, and consistency issues often belong to different owners.
• Write a small prompt panel: keep 10 to 20 prompts covering branded, non-branded, use-case, and comparison intent.
• Re-check after shipping: do not wait for a quarterly review. Re-test after meaningful changes so you can tie outcomes to actions.

This is also where execution usually breaks down. Teams see the problem but do not convert it into a backlog. BotRank's Recommendations feature is useful here because it turns audit findings into concrete GEO tasks instead of leaving them as a loose list of observations.

FAQ: what do teams usually get wrong in a free AI visibility audit?

Is one chatbot test enough to judge AI visibility?

No. One answer can be a fluke. Use a small prompt set across multiple engines and compare the results.

Do I need structured data to be visible in AI search?

Not as a magic switch, but it helps machines understand the entity and the page more reliably. Think of schema as clarity infrastructure, not a shortcut.

What is the fastest free fix for many brands?

Fact consistency. Cleaning up mismatched business details, vague product descriptions, and unsupported claims often improves trust faster than publishing another generic article.

When should a brand move from manual checks to a platform?

As soon as the team needs trend lines, stable prompt monitoring, competitor benchmarking, or cited-source analysis. Manual audits are strong for diagnosis and weak for operational measurement.

A free AI visibility audit is not a replacement for ongoing tracking. It is the right first step when you need to stop guessing. Run the nine checks, fix the obvious gaps, and then measure whether your brand becomes easier for AI systems to find, understand, and recommend.