Safari MCP for AI debugging, seoClarity's markdown guide, agents skipping hard-to-parse pages, and llms.txt checks.

Visibility #24: Serve Markdown to AI Bots Now

Week of June 30 - July 6, 2026 | AI Search & Visibility Intelligence

It was a quieter news week for big platform announcements, but the technical layer got more interesting. Apple shipped an MCP server for Safari that opens AI-based debugging for SEO and Core Web Vitals. seoClarity published a practical guide to serving markdown to AI bots. And the "what agents skip" question finally has some data behind it. Let's get into it.

Safari's New MCP Server Opens AI Debugging for SEO

Apple's WebKit team shipped an MCP (Model Context Protocol) server for Safari that lets AI agents directly debug websites - including SEO signals and Core Web Vitals. In practical terms, this means developers and SEOs can now route Claude, Cursor, or other MCP-compatible agents through Safari's diagnostic layer to identify rendering issues, CWV bottlenecks, and crawlability problems without leaving their AI toolchain. We've been watching Apple's AI search moves closely; this feels like infrastructure being laid quietly before a bigger shift. An AI search default in Safari would reshape referral traffic for every publisher with an organic strategy - and building debugging tools into the browser stack looks like groundwork for exactly that.

📎 Search Engine Journal​

seoClarity: How to Serve Markdown to AI Bots

seoClarity published a technical guide on serving markdown directly to AI agents - crawlers like Claude Code, OpenAI's Operator, and others that increasingly prefer clean markdown over HTML-heavy pages. The core argument: AI agents parsing structured markdown get cleaner signal, cite more reliably, and are less likely to skip your content due to rendering complexity. The guide covers conditional serving (markdown for bots, HTML for humans), implementation via middleware, and what the agentic web means for technical SEO priorities. This is the kind of thing that probably seems optional today and won't in twelve months.

📎 seoClarity Blog​

Where Clicks Go, What Agents Skip

Search Engine Journal's SEO Pulse piece this week covers three things worth separating out: where clicks are actually going in AI-mediated search, which content types AI agents skip when crawling for citations, and who is leaving Bing and why. The agents-skip finding is the most actionable - content behind login walls, JavaScript-heavy pages, and thin pages with low information density are being passed over at higher rates. This connects directly to seoClarity's markdown guide above: if agents can't easily parse your content, they won't cite it. The Bing data is worth watching as a leading indicator of AI search consolidation.

📎 Search Engine Journal​

Google Put AI Visibility Inside Search Console on Purpose

SEJ's piece on Google's AI visibility integration in Search Console argues this wasn't an accidental product decision - Google is deliberately positioning Search Console as the measurement layer for AI Overviews and AI search performance, not just traditional organic. The implication: Google wants to be the platform where you track AI visibility, which influences what data they surface and what they don't. In our experience, treating any single platform's native analytics as the full picture is how you end up with a blind spot. Search Console AI data is a useful signal; it's not the whole story.

📎 Search Engine Journal​

Lighthouse Now Fails llms.txt Without Markdown Links

Lighthouse - Google's site quality auditing tool - is flagging llms.txt files that don't include properly formatted markdown links. This matters because llms.txt is increasingly used to communicate site structure and content permissions to AI crawlers. If your llms.txt uses plain URLs instead of markdown-formatted links, you're now getting a Lighthouse failure. Small technical issue, but it reflects how fast the standards for AI-crawler-readable content are solidifying. Check your llms.txt formatting if you've deployed one.

📎 Search Engine Journal​

So What Do You Do About It?

The through-line this week is the technical layer of AI search visibility. Safari's MCP server, seoClarity's markdown guide, and the Lighthouse llms.txt update all point the same direction: AI agents are increasingly picky about how they consume content, and the infrastructure expectations are hardening faster than most content teams have noticed.

Serving clean markdown conditionally - to bots, not humans - is probably the highest-leverage technical move you're not making yet. It's not complicated to implement, and the agent-citation upside is real. Pair that with a current llms.txt audit and you've covered the most actionable technical ground this week surfaced.

One thing this week: Pull up your llms.txt (if you have one) and run it through Lighthouse. If you don't have one yet, that's the prior step. Then check whether your highest-value content pages are JavaScript-heavy or login-gated - those are the pages agents are currently skipping.

More From This Week

From the Tool Blogs

• ​Fuel the Agentic Web: How to Serve Markdown to AI Bots (seoClarity) - Technical implementation guide for conditional markdown serving; covered in depth above, but worth bookmarking as a reference.
• ​100 Most Expensive Keywords for Google Ads (July 2026) (Ahrefs) - Updated CPC data. Mesothelioma and legal keywords still dominate; useful for benchmarking competitive PPC landscapes, less directly GEO-relevant but strong as a data reference.
• ​100 Most Asked Questions on Google (July 2026) (Ahrefs) - Updated question volume data. Cross-reference against your content's FAQ coverage to identify AI Overview answer opportunities.
• ​Botify: Agentic Catalog vs PIM vs Feed Management (Botify) - Positioning piece on how agentic content catalogs differ from traditional product information management; relevant for e-commerce teams thinking about AI crawlability.
• ​Botify: Fast Revenue Impact in the AI Era (Botify) - Framework for connecting technical SEO improvements to revenue outcomes in an AI-mediated search environment.

Agency Insights

• ​What Is Query Fan-Out in AI Search? (Search Influence) - Clear explainer on how AI search engines expand a single query into multiple sub-queries. Understand this and your content strategy for AI visibility makes more sense.
• ​What Is Information Gain in SEO? (Search Influence) - The concept Google uses to evaluate whether your content adds something net-new vs. restating existing answers. Directly tied to AI citation likelihood.
• ​We Didn't Automate Our Content. We Automated the Tedium. (Search Influence) - Practical account of where automation helps in content production for AI search and where it doesn't. More honest than most agency content on this topic.
• ​Short-Form Video Is Showing Up in B2B Search Results and AI Answers (Foundation Inc) - Data on short-form video appearing in B2B AI answers. If your GEO strategy doesn't include video, this is a gap worth examining.
• ​Google Just Published How It Catches AI Content at Scale (Foundation Inc) - Breakdown of Google's public documentation on AI content detection. What happens when it flags your content, and what the signals are.
• ​How a Branded Subreddit Helps Cursor Own BoFu Discussions (Foundation Inc) - Case study on how Cursor uses a branded Reddit community to dominate bottom-of-funnel AI citations. Concrete example of the entity-building strategy we've been covering.

The Visibility Report | Will Scott​
This newsletter is produced collaboratively by Will Scott and Bob, an AI agent. Human oversight, AI efficiency.
​Unsubscribe​