Plain-text Lifelines: How to Make Medical Content Visible in AI Search

Today, a lot of potentially life-saving medical content is hiding in plain sight. If you’ve completed the latest module in Sitebulb’s JavaScript SEO training course, you’ll know why.

Most LLM bots don’t render JavaScript.

They fetch HTML, strip scripts, toss out CSS and JSON, and read whatever text is left.

As Dan Taylor put it in our recent webinar (with Joshua Boughton), “Most LLMs don't render JavaScript content… they look at the raw HTML… It's more like a text-based scraper.”

That’s fine if your condition pages are server-rendered. But it’s a problem if dosage tables, symptom lists and FAQs sit inside fancy accordions that only appear post-render.

So if your organisation – whether a charity, NHS trust or private clinic – relies on JS-heavy components to deliver clinical guidance, there’s a real risk AI search won’t surface your advice at the moment people need it.

Who renders what

It's not all bad news.

Some AI surfaces sit on search engine infrastructure that can render (at least sometimes); here’s looking at you Gemini.

In Dan’s words: “AI Overviews and AI Mode… technically have rendering capabilities… ChatGPT is… inconclusive; sometimes it did, sometimes it didn't.”

Translation? Don’t bank on render always happening. Design for a world where your content must be readable as raw HTML, first time, every time.

Give crawlers the “plain-text” version of your story

Here’s a simple trick that Dan shared: give AI crawlers a lightweight, easy-to-read version of your pages. Basically, a copy without all the fancy layout scripts.

Think of it like a clinician’s briefing sheet: clean headings, bullet points, tables and summaries, minus the pop-ups, widgets and interactive bits. All the essential information, none of the fluff.

If you want to do this safely, especially for medical sites:

• Keep it official: Your plain-text version should pull from the exact same approved content as your main website, not a separate or manually-edited copy. That way, you won’t end up with two slightly different versions of the same advice floating around online.

• Keep it current: Medical guidance changes. Make sure this version updates whenever your main page does, so out-of-date information doesn’t linger.

• Share it with the right bots: Only show this stripped-down view to known AI crawlers (like PerplexityBot or ClaudeBot). Everyone else should still get the normal, nicely formatted page.

You can double-check all of this in Sitebulb. Its single-page analysis lets you pretend to be different crawlers and see exactly what each one receives. Then, run a quick crawl to make sure those “plain-text” pages are actually connected and findable on your site.

Schema is hygiene, not magic

You’ve probably seen the chat about LLMs ignoring schema. Well, here’s the slightly nuanced version from Dan:

“LLMs can understand structured data… they just don't often receive it… as plain text or markdown when it gets through to processing.”

Dan Taylor   Head of Technical SEO, SALT.agency

In other words, don’t expect MedicalCondition markup to switch on LLM visibility by itself.

Do implement granular healthcare schema to help the broader search ecosystem, improve eligibility for rich results, and reinforce knowledge-graph context.

Practicalities for medical teams:

Don’t feel you need to mark up everything on your site. Just focus on the pages where schema genuinely adds clarity.

• Be specific: Use schema types that actually describe what’s on the page, e.g. MedicalCondition for illnesses, Drug for medication information, or MedicalWebPage for general advice articles. The goal is to help search engines clearly understand what kind of medical content they’re looking at.

• Show your credentials: Make sure your author bios, citations, and references are solid. In Google’s eyes, medical information is a YMYL topic, so E-E-A-T matters more than ever.

• Keep it maintained: Schema standards change over time, and old types sometimes get retired. Run regular checks to make sure your markup is still valid and up to date. Sitebulb can help here: You can sign up for Sitebulb’s structured data alerts to be notified of any changes.

Robots.txt: block or allow

Another easy way to shoot yourself in the foot is robots.txt.

For healthcare organisations, it’s a delicate balance between public good and data governance.

Action points:

• Decide policy intentionally: If you want inclusion in AI answer surfaces, don’t block every LLM agent blindly.

• Understand CCBot: Blocking CCBot may keep you out of some training sets, but it may also reduce beneficial discoverability routes.

• Audit regularly: Use logs, third-party monitors and a Sitebulb crawl to confirm your directives match reality.

Speed vs vitals: what actually correlates

Here’s a little relief for anyone who’s been losing sleep over PageSpeed scores. As Dan put it, “Most websites being cited in LLMs aren’t lightning fast.”

So you don’t need to chase a perfect 100 on Lighthouse. What really matters are your LCP (Largest Contentful Paint) and CLS (Cumulative Layout Shift) scores.

For medical pages that tend to be heavy with charts, tables or images, try these small but meaningful tweaks:

• Tame your main image: Generate your “hero” image on the server and make sure it has a set width and height. That stops the whole page from jumping around as it loads.

• Prioritise the important bits: Load your most important content – like the first paragraph or symptom summary – straight away. You can do this by keeping the critical styling inline so it doesn’t have to wait for everything else to download.

• Cut the clutter: Ditch decorative icon fonts or bloated scripts and use simple system icons instead. Lighter pages mean faster, more stable experiences.

• Don’t hide your main content behind scripts: Make sure crucial information – like treatment details or dosage guidance – appears instantly in the HTML.

URL myths, stable anchors, and highlightable snippets

When it comes to URLs, you can relax; you don’t need to start chopping them down to five words or less. Dan ran a study at SALT.agency, looking at more than 100,000 URLs from both traditional Google search results and AI-generated answers (specifically Google’s AI Overviews and AI Mode).

The verdict?

“URL length doesn’t have any difference in citation.”

In other words, AI tools aren’t favouring shorter links. A long, descriptive URL like /treatment/what-to-expect-in-alcohol-detox is just as likely to be cited as a snappy /detox-guide. What does help, though, is how your content is structured within the page itself.

AI systems like Google’s AI Overviews sometimes link directly to a highlighted line of text using #:~:text. Dan checked where these highlights appeared on hundreds of pages and found that the “pixel depth” – how far down the page the quote came from – varied wildly. In other words, AI isn’t just grabbing the first paragraph every time.

So what should you do?

• Keep your sections clear and scannable, with logical headings.

• Place key takeaways or definitions near the top, so they’re easy to find.

• Avoid constantly changing your headings or anchor IDs; stability helps AI (and humans) link directly to the right bit of text over time.

It’s less about crafting “perfect” URLs and more about making your pages easy to quote accurately, no matter who’s doing the quoting.

Fundamentals first (and forever)

If you only take one mantra back to your team, borrow Joshua’s: “You can't cook in a dirty kitchen.” So get your house in order. That means periodic content audits, retiring outdated guidance, consolidating duplicative topics and redirecting legacy URLs.

Joshua also reminded us that bots behave like tireless readers: “Internal linking… bots are just clicking around… put the food on the plate for them [and make it easy for them] to digest.”

• See how deep your key pages are buried: Sitebulb can map out your site’s internal linking and show how many clicks it takes to reach each page. If your core condition or treatment pages are hidden behind too many layers, bring them closer to the surface by linking to them directly from your main navigation or key content hubs.

• Create a logical journey: Guide both users and crawlers through a clear flow, e.g. symptom → condition → diagnosis → treatment. Use descriptive headings and anchor links so it’s obvious what each section covers and easy to move between them.

• Clean up the dead ends: Remove or redirect orphan pages and thin content that don’t add real value. They only distract crawlers and dilute the authority of your stronger pages. Learn more: How to find and fix orphaned pages

Redefine ‘winning’ when clicks don’t happen

In many AI journeys, the answer happens off-site. Joshua’s advice:

“Expand your definition of SEO success… own as much real estate in the SERP as you can.”

Joshua Boughton   Senior SEO Manager, BloomHouse Marketing

Look at a slightly different mix of signals to understand your impact. Here’s how to build that fuller picture:

• Google Search Console impressions: Tracking impressions for your key medical topics gives you a sense of your ongoing visibility; especially useful if you’re trying to measure brand reach rather than raw traffic.

• Referral traffic analysis: Some AI tools (and other aggregators) still pass along a referral source when people click from their results to your site. Keep an eye on your landing pages in Google Analytics to see if new traffic sources pop up.

• LLM tracking data: There are plenty of LLM tracking tools like Waikay, Peec AI and Profound, so you can see where your brand or website is being cited inside AI-generated answers. It’s an emerging space so not perfect yet, but it helps you understand how your information is being surfaced beyond traditional search.

• Real-world outcomes: Finally, tie all this back to what actually matters for your organisation: are helpline calls increasing? Are more users starting online assessments or booking appointments? These are the real indicators that your content is doing its job, even if AI is the middleman.

A short, sustainable playbook for medical orgs

1. Render for text first. SSR by default; add a clean-text fallback for trusted bots.

2. Strengthen E-E-A-T. Authors, citations, review processes.

3. Implement schema as hygiene. Accurate, granular, validated.

4. Tidy the kitchen. Audit, deprecate, redirect; keep content current.

5. Fix vitals that matter. Prioritise LCP/CLS on core advice pages.

6. Link like a clinician thinks. Clear paths through the care journey.

7. Measure visibility differently. Impressions, referrals, citations, outcomes.

Do that, and you won’t just be “optimising for AI.” You’ll be doing what good medical publishers have always tried to do: making trustworthy answers easy to find, regardless of where the question gets asked.