Skip to content
Agent Checker
Back to Blog
TechnicalSEOAI Agents

Open Graph and Meta Tags: What Agents Read Before They Visit

Agent Checker4 min read

Before an AI agent loads your page, it reads the metadata. The og:title, og:description, and standard meta description act as a preview layer, giving the agent enough context to decide whether the page is worth visiting. Get these wrong, and agents may skip your content entirely.

The preview layer agents rely on

When an agent encounters a link to your site, whether in search results, a social feed, or another page, it often has access to Open Graph tags and meta descriptions before it renders anything. These tags serve as a summary. The agent reads them, decides if the page matches the user's intent, and either visits or moves on.

This is similar to how search engines use meta descriptions in snippets, but agents take it further. They use the metadata to build an internal model of what the page contains before committing resources to load it.

Which meta tags matter most

Not all meta tags carry equal weight. Here is what agents actually read:

<head>
  <title>Wireless Keyboard | TypeWell UK</title>
  <meta name="description" content="Compact wireless keyboard with UK layout. Bluetooth 5.0, 6-month battery life. £49.99 with free delivery." />
  <meta property="og:title" content="Wireless Keyboard | TypeWell UK" />
  <meta property="og:description" content="Compact wireless keyboard with UK layout. Bluetooth 5.0, 6-month battery life." />
  <meta property="og:type" content="product" />
  <meta property="og:url" content="https://example.com/keyboards/wireless" />
  <meta property="og:image" content="https://example.com/images/keyboard-hero.jpg" />
</head>

meta description is the most useful tag for agents. It should describe what the page offers in concrete terms, with specifics like prices, quantities, or key features.

og:title and og:description often duplicate the title and meta description, but agents may prefer them because the Open Graph protocol is more structured. If they differ from the standard tags, agents may get confused about which version to trust.

og:type tells the agent what kind of content to expect: article, product, website, profile. This helps agents categorise your page before loading it.

og:image matters less for text-based agents, but visual agents that render previews will try to fetch this image. Broken image URLs create a poor first impression.

Common mistakes that trip up agents

Missing descriptions entirely. A surprising number of pages have no meta description at all. The agent gets a title and nothing else. It has to guess what the page contains, and guessing wastes time.

Generic, templated descriptions. "Welcome to our website. We offer the best products and services." This tells an agent nothing. Be specific. What products? What services? What makes this page different from your other pages?

Mismatched metadata. When your og:title says "Summer Sale" but your <title> tag says "Product Catalogue", agents have to reconcile the conflict. Keep them consistent.

Outdated og:image URLs. If you redesigned your site six months ago but the Open Graph images still point to old CDN paths, agents that try to fetch previews will hit 404 errors. Check these periodically.

Stuffing keywords. "Best cheap wireless keyboard UK buy online discount sale keyboard wireless." Agents are sophisticated enough to recognise keyword stuffing, and it degrades the signal they extract from your metadata.

How agents use metadata differently from search engines

Search engines use meta descriptions primarily for display in results pages. Agents use them for decision-making. An agent working on a task like "find me a keyboard under £50" will scan meta descriptions across multiple pages, filter based on the content, and only visit the pages that seem relevant.

This means your meta description is not just a marketing snippet. It is functional data. The more precise and accurate it is, the more likely an agent will visit your page for the right reasons.

Structured data as a complement

Meta tags work well alongside Schema.org markup to give agents a complete picture. Where Open Graph tags provide a quick preview, structured data like JSON-LD provides the deep detail agents need once they arrive on the page.

Think of it as two layers: meta tags are the shop window, and structured data is the product catalogue inside. Both matter, but the shop window determines whether anyone walks in.

Practical steps

  1. Audit every page template for missing meta description and og:description tags
  2. Write descriptions that include specific, factual details about the page content
  3. Keep og:title and <title> consistent
  4. Verify og:image URLs return valid images (not 404s)
  5. Set og:type appropriately for each page type

You can run an agent readiness audit to see exactly how agents experience your metadata. The results often reveal gaps that are invisible during normal browsing but obvious to an automated visitor.