This Free Tool Shows What AI Actually Sees When It Visits Your Website. It's Not Pretty.

I owe Benjamin Tannenbaum an apology.

He posted on LinkedIn about a free tool he'd built, an "AI browser" that shows you what AI systems actually see when they visit your website. I saw it, got excited, and immediately left comments that were basically: "Great post! By the way, HERE'S OUR AI VISIBILITY SERVICE." The kind of drive-by self-promotion that makes LinkedIn insufferable. I deleted them, but not before I'd already embarrassed myself.

So, Ben: sorry about that. Passion got the better of me.

His point deserves proper attention. Not a sales pitch masquerading as engagement, but an honest look at the problem his tool reveals. Because what it shows is something most business owners have never considered: the website you spent thousands on looks completely different to an AI than it does to you.

Try It Yourself (Seriously, Do This Now)

Ben's tool is called the AI Content Reader, part of his company GetAISO. It's built on md-browse, an open-source Markdown viewer that converts web pages into the clean, text-only format that AI systems work with.

Here's what you do: paste your website URL into it. Then look at what comes back.

That gorgeous hero section with the full-width photography? Gone. Your carefully chosen brand colours? Invisible. The animated counter showing "500+ happy clients"? Doesn't exist. The interactive pricing table that toggles between monthly and annual? Vanished.

What's left is text. Just text. With some basic structure (headings, links, lists) if you're lucky. If your site relies heavily on JavaScript to load content, there might be even less than that.

As Ben put it in his LinkedIn post: "Humans see a page as a visual experience. Layout, menus, tabs, accordions, carousels, popups, infinite scroll, images, embedded widgets. AI search assistants usually do not 'see' that experience."

He's right. And the gap between what you see and what AI sees is bigger than most people realise.

What AI Actually Sees When It Visits Your Site

When ChatGPT, Claude, Perplexity, or Google's AI Overviews visit your website, they don't open a browser. They don't render your CSS. They don't execute your JavaScript (with some exceptions). They read the raw HTML and extract text.

Think of it as reading the blueprint instead of walking through the building. Everything that makes your office look professional, the paint, the furniture, the reception area, all of that disappears. What's left is walls and doors. If your business card is pinned to a noticeboard inside a locked room (read: hidden behind a JavaScript accordion), AI never finds it.

Specifically, here's what disappears:

What AI crawlers can and cannot see on your website

Element	Humans See	AI Sees
CSS styling and layout	Beautiful design, brand colours, spacing	Nothing. Invisible.
Images without alt text	Photography, graphics, illustrations	Blank space. No information.
JavaScript-loaded content	Reviews, pricing, dynamic elements	Often missing entirely
Accordions and tabs	Click to reveal content	Hidden or collapsed text
Image carousels	Sliding gallery of images	Maybe the first image's alt text
Interactive calculators	Quote builders, ROI tools	Empty form elements
Video content	Embedded YouTube, explainer videos	An iframe tag. No content.

The JavaScript Blind Spot

Here's where it gets properly concerning. Google's crawler, Googlebot, has spent over a decade learning to render JavaScript. It does a decent job. But most AI crawlers? They're simpler. GPTBot, ClaudeBot, PerplexityBot: they tend to read the initial HTML response and move on. If your content loads after a JavaScript event (a click, a scroll, a timer), these crawlers never see it.

That matters more than you'd think. Practically every modern website uses JavaScript. If your testimonials load via a third-party widget, AI doesn't see them. If your FAQ section is an accordion that collapses content until clicked, AI might miss it. If your pricing sits behind a "view plans" toggle, it's gone.

"Gen AIs value fresh content in particular, partly as a reference check of their LLM training data."

Fabrice Canel, Principal Product Manager at Microsoft Bing, SMX Munich 2025

When I first read that quote, it clicked. AI systems aren't just scraping your site once and filing it away. They're actively checking back, looking for fresh information to validate what they already know. If your content is locked behind JavaScript interactions, they come back, find nothing new, and move on to a competitor whose content is readable.

Canel also confirmed at the same conference that schema markup helps Microsoft's LLMs understand content. Google hasn't made the same confirmation publicly, but if Bing's AI uses structured data, it would be surprising if Google's Gemini-powered AI Overviews ignored it entirely.

The Problem Ben's Tool Doesn't Solve

The AI Content Reader is brilliant at showing you the problem. Paste your URL, see the stripped-down version, understand the gap. Go try it. Honestly. I'll wait.

But seeing the problem is step one. Even if your website has clean HTML, proper heading structure, descriptive alt text on every image, and no JavaScript-dependent content, there's still a deeper issue: your website doesn't tell AI who you are.

HTML tells AI what's on the page. It doesn't tell AI what your business does, where you're based, what industries you serve, how to pronounce your brand name, or what you'd prefer it says when recommending you. Your about page might contain some of this, but AI has to infer it from natural language written for humans, not machines.

That's like handing someone your novel and expecting them to write your CV from it. The information might be in there somewhere, but it's not structured for that purpose.

AI Discovery Files: The Layer Your Website Is Missing

This is where the conversation goes beyond what Ben's tool shows. AI discovery files are a set of machine-readable files you place at the root of your website that explicitly tell AI systems who you are, what you do, and how to represent you.

They work alongside your website, not instead of it. Google's John Mueller called markdown pages for AI "a stupid idea" when people started creating duplicate content in plain text. He was right about that. But AI discovery files aren't content mirrors. They're structured metadata, closer to robots.txt or sitemap.xml than to a duplicate of your homepage.

Here's what each file does:

Core AI discovery files and their purpose

File	What It Tells AI
llms.txt	A plain-text summary of your business for large language models: who you are, what you do, your key services, and links to important pages
ai.json	Structured data about your business in JSON format: service areas, team, pricing model, USPs
identity.json	Your canonical business identity: official name, pronunciation, founding date, location, and how to represent you
brand.txt	Brand representation rules: how to spell your name, what to call you, what not to say about you
faq-ai.txt	Verified Q&A pairs that AI can use directly when answering questions about your business

We've seen the difference first hand. When we asked Google Gemini what it learns from our AI discovery files, it identified five categories of information it couldn't extract from our website alone: brand naming rules, verified Q&A pairs, service boundaries, founder context, and pricing model structure.

Companies like Anthropic, Cloudflare, Stripe, and Vercel already use llms.txt files. The adoption curve is steeping fast. With Anthropic's $30 billion funding round accelerating Claude's reach, the number of AI systems reading your website is only going up. But for UK small businesses, it's still early days. When we tested 100 UK small businesses in ChatGPT, only 3 had any AI discovery files at all.

Three out of a hundred. That's the current state of play.

Check Your Own AI Visibility

Ben's tool shows you what AI sees when it reads your HTML. Our AI Visibility Checker goes a step further: it actually asks AI systems about your business and shows you what they say.

The checker runs live queries against ChatGPT and shows you the real responses. Not a simulated score. Not an estimate based on your metadata. The actual words ChatGPT uses when someone asks about your business. We call it the live AI snapshot, and it's the difference between checkers that guess and checkers that know.

"59.7% of EU Google searches and 58.5% of American Google searches resulted in zero clicks."

Rand Fishkin, Co-founder of SparkToro, 2024 Zero-Click Search Study

I've been building websites for over thirty years. For most of that time, the game was clear: rank on Google, get clicks, convert visitors. Fishkin's data shows that game is dying. Almost 60% of searches now end without anyone clicking through to a website. AI answers the question directly. If the AI's answer about your business is wrong, incomplete, or nonexistent, you've lost that customer before they ever saw your site. And with ChatGPT ads now live, that organic answer sitting above the paid placements carries even more weight.

And here's the thing: traditional SEO won't fix this. SEO optimises for keywords and rankings. AI visibility optimises for understanding. The two overlap, but they're not the same thing. You can rank first on Google and still be invisible to ChatGPT.

What To Do About It

Start with Ben's tool. Seriously. Paste your URL into the AI Content Reader and see what comes back. If the result looks thin, sparse, or missing key information about your business, you've got work to do.

Here's the priority list, from quick wins to proper overhaul:

Quick Wins (This Afternoon)

• Alt text on every image. Not "image1.jpg" or "hero banner". Describe what's in the image. AI reads alt text when it can't see the picture.
• Semantic HTML. Use <h1> through <h6> for headings, not styled <div> tags. Use <article>, <section>, <nav> instead of generic containers. This gives AI the structural clues it needs.
• Move critical content out of accordions. If something matters, put it in the open. Collapsed content is a gamble with AI crawlers.

Medium Term (This Week)

• Create an llms.txt file. A plain-text summary of your business at yoursite.com/llms.txt. The spec takes ten minutes to follow. WordPress users can use our AI Discovery Files plugin to generate it automatically.
• Add Schema.org markup. LocalBusiness, Organization, FAQPage: the structured data that Bing has confirmed helps its LLMs understand your content. Google hasn't said it explicitly, but the logic applies.
• Run the AI Visibility Checker. See exactly what ChatGPT says about your business today. The live snapshot will either reassure you or light a fire.

Long Term (This Month)

• Implement the full AI discovery file suite. llms.txt, ai.json, identity.json, brand.txt, faq-ai.txt. Full specs are on the ai-visibility.org.uk site.
• Audit your JavaScript dependencies. If key content loads via JS, ensure it's also in the initial HTML response. Server-side rendering or progressive enhancement fixes this.
• Review your AI visibility quarterly. AI systems update their training data and crawling behaviour regularly. What works today might need adjusting in three months.

Give Ben Your Feedback

If you try the AI Content Reader and find it useful (or find ways it could improve), let Benjamin know on LinkedIn. Constructive feedback only, please. The tool is free, it's solving a real problem, and the man deserves better than what I gave him with my first reaction.

The gap between what humans see and what AI sees on your website isn't going to close on its own. Tools like Ben's make the problem visible. AI discovery files and proper AI optimisation are how you fix it. And the sooner you start, the better, because right now, 97% of UK small businesses haven't.

Frequently Asked Questions