A crawler sees your page. AI needs the fields behind it.
Much of what AI uses to describe a business lives in CMS fields that never render cleanly on the page. That’s where to look first.
The page is only part of the picture
When an AI engine describes your business, it reads more than the words a visitor sees. Titles, meta descriptions, structured data, image alt text, and the way the content is organized ship with every page, and together they tell a machine what the page is about.
In March 2025, Fabrice Canel, a principal product manager for Microsoft Bing, was reported to have confirmed that schema markup helps Microsoft's LLMs, including Copilot, understand content. Google is more measured. Its guidance on AI features and your website says a page needs to be indexed and eligible for a snippet, that there are no additional technical requirements, and that structured data should match the visible text.
So the goal is not more markup. It is accurate markup, on the pages that matter.
What AI can read is what ships
Most of that machine-readable layer is produced by your CMS and templates from fields an editor filled in. AI crawlers read the HTML your server returns. In Vercel and MERJ's 2024 measurements, none of the major AI crawlers rendered JavaScript, while Gemini, which uses Googlebot's infrastructure, did. A fact that only appears after a script runs, or that sits in a CMS field no template outputs, is a fact those engines never see.
A standard crawler sees the rendered result but not the reason: which field is empty, which template drops it, which value was edited on one page and left stale on another. An audit built on crawling can miss why a page is misread.
Where the gaps usually are
- Schema fields left blank in page templates.
- Meta descriptions left on auto-generated defaults. Google says it builds snippets primarily from page content and uses the meta description when it describes the page better, and a generic default rarely does.
- Events, locations, or products stored as free text instead of structured fields.
- Facts in the CMS that no longer match what’s published, including structured data that disagrees with the visible text.
A designer's view: the edit screen is part of the problem
These gaps are rarely an editor's mistake. They are interface decisions. The edit screen is where the machine-readable half of a page gets written, and most edit screens never mention that a machine will read it. A few changes, all of them design work and none of them a new platform:
- Say who the field is for. "Meta description" tells an editor nothing. "Shown in search results and read by AI assistants: one or two plain sentences" tells them what a good answer looks like.
- Use the input that fits the fact. Opening hours, addresses, prices, and event dates belong in structured components with validation, not in a text box. Free text can't be checked, and it can't be turned into markup.
- Make the important fields required. Optional fields are empty fields by default.
- Show what ships. A preview of the snippet and structured data beside the form makes the invisible half of the page visible while the editor is still working.
- Check on save. If a structured value disagrees with the visible text, say so before the page is published, not months later.
- Treat alt text as content. It serves people using screen readers and gives machines a plain-text description of the image. Google lists supporting your text with high-quality images among the fundamentals that still apply to its AI features.
None of this is specific to AI. It is the same discipline as designing a good form for a person, applied to a second reader.
What to check first
Start with the pages that answer buying questions: services, pricing, locations, and anything a customer would ask AI to compare. Check whether their structured fields are filled, specific, and consistent with what the page says. Fix those before you rewrite a single headline.
The content model behind a page decides what AI can reliably say about you.
How TROVR approaches it
TROVR connects to your CMS and reads the fields behind each page alongside the page itself. It proposes field-level changes ranked by impact, and once you approve them, writes them back into the CMS you already use. See how it works.
Sources
- AI features and your website, Google Search Central, last updated December 2025.
- How to write meta descriptions, Google Search Central.
- The rise of the AI crawler, Vercel and MERJ, 2024.
- Schema helps Microsoft's LLMs (Copilot) understand your content, Search Engine Roundtable, March 2025.
Frequently asked questions
Do I need special markup to appear in Google's AI Overviews or AI Mode?
No. Google says a page needs to be indexed and eligible to show in Search with a snippet, and that there are no additional technical requirements. It also says any structured data you use should match the visible text on the page.
Do AI crawlers run JavaScript?
In Vercel and MERJ's 2024 measurements, none of the major AI crawlers they tested rendered JavaScript, including OpenAI's, Anthropic's and Perplexity's. Google's Gemini is the exception because it uses Googlebot's rendering. Content that only appears after a script runs may not be seen, so ship the important facts in the HTML.
Does schema markup help AI understand a page?
Microsoft's Fabrice Canel, a principal product manager for Bing, was reported in March 2025 to have confirmed that Microsoft's LLMs use schema markup to understand content, according to Search Engine Roundtable. Google's guidance is narrower, so treat markup as a way to state facts clearly, not as a shortcut to being cited.

