AI Citation Share Declines; New Data Questions LLMS.txt

Welcome to Pulse. This week’s developments center on how to measure AI visibility, what to expect from llms.txt, and where the agentic web is heading next.

Microsoft introduced new AI citation tools, fresh data questioned the value of llms.txt, Google and two coalitions released agent specifications, and the UK set new ranking rules for Google Search.

Here’s what matters for you and your work.

—

Bing Rolls Out AI Citation Share In Webmaster Tools

Microsoft is adding four features to the Bing Webmaster Tools AI Performance dashboard: Citation Share, Intents, Topics, and Compare. All are currently in preview.

Key facts: Citation Share shows the percentage of AI citations your site captures for a given grounding query. Intents and Topics group those queries to work around a data limit in the current dashboard. Compare lets you overlay a past period on the current one. These features are rolling out globally, still in preview.

Why This Matters

Citation Share is the first Bing Webmaster Tools metric that reveals your AI visibility relative to competitors, not just whether you were cited. The limitation is that it only covers Bing data, including Copilot and Bing’s own answers. It says nothing about Google, where Search Console still lacks any citation-style counts.

What SEO Professionals Are Saying

Gianluca Fiorelli, Founder of ILoveSEO.net, wrote on LinkedIn: “Bing Webmaster! The Google Search Console we would like to have.”

Read our full coverage: Bing Rolls Out AI Citation Share In Webmaster Tools

—

Google And Ahrefs Data Narrow The Case For llms.txt

llms.txt took two hits this week. Google’s John Mueller said the file cannot help an LLM distinguish one site from another. New Ahrefs data showed the bots that matter barely fetch it.

Key facts: Speaking on Search Off the Record, Mueller argued that llms.txt cannot differentiate sites for discovery because the file is self-reported by the site hoping to be chosen. He pointed back to ordinary HTML and internal links instead. The Ahrefs data supports this. Across 137,000 domains, 97% of llms.txt files received zero requests. The retrieval bots that generate citations, like ChatGPT and Perplexity, accounted for just 1% of the fetches that occurred.

Why This Matters

Both findings point in the same direction. A self-reported file cannot make an LLM choose you, and the bots that generate citations barely fetch it. Do not expect llms.txt to move your AI search visibility. It still has a narrow role with coding agents and training crawlers that read it, making it cheap to maintain. This echoes the conclusion SE Ranking reached months ago after examining 300,000 domains.

What SEO Professionals Are Saying

Nat Miletic, Founder at Clio Websites, summed it up on LinkedIn: “llms.txt is low cost to publish, fine to have. Just don’t expect it to move AI visibility right now.”

Read our full coverage: Google’s Mueller Says llms.txt Can’t Help LLMs Differentiate Sites and 97% Of llms.txt Files Got No Requests, Ahrefs Data Shows

—

Google, Microsoft, And Others Publish New AI Agent Specs

Two agent specifications landed within days of each other. Google Cloud published the Open Knowledge Format, and a coalition including Google, Microsoft, GitHub, and Hugging Face followed with Agentic Resource Discovery.

Key facts: The Open Knowledge Format (OKF) is a markdown format for packaging organizational knowledge like datasets, metrics, and runbooks in a way AI agents can read. Agentic Resource Discovery (ARD) is a draft spec for how agents find and verify tools, skills, and other agents. Both are early, with OKF at version 0.1 and ARD at 0.9.

Why This Matters

Neither spec requires action from you this week. Both repeat the move llms.txt made: a structured file on your own domain for software to read. The same unsettled question of adoption hangs over them. Watch which formats catch on before committing to any.

What SEO Professionals Are Saying

Martin Jeffrey, Founder and Strategic Lead at Harton Works, compared ARD to the early days of search on LinkedIn: “It is the sitemap, reborn for capabilities rather than pages.”

Suganthan Mohanadasan, Co-founder at Snippet Digital, who built two free tools for the format, tempered expectations: “This is not a magic mushroom and won’t increase your AI visibility overnight.”

Read our full coverage: Google Cloud Announces The Open Knowledge Format and Google, Microsoft Back Draft AI Agent Discovery Spec

—

UK Orders Google To Rank Search Results Fairly

The UK’s Competition and Markets Authority (CMA) set new rules for Google Search. Google must rank organic results on objective criteria and give notice before significant changes.

Key facts: The Fair Ranking requirement covers UK organic results, including AI Overviews but not ads. It obliges Google to use objective, non-discriminatory criteria, provide advance notice of significant changes, and offer a route to raise ranking concerns. A second requirement turns Google’s voluntary UK data portability tool into a legal obligation. Google disputed the premise, saying its ranking is already fair and transparent.

Why This Matters

The advance-notice and complaints requirements could affect daily work. They replace the black-box core update with a warning and a way to push back. Fair ranking also applies inside AI Overviews. Like the CMA’s early-June opt-out requirement, it applies only in the UK. The real impact depends on how Google chooses to implement it.

What SEO Professionals Are Saying

Laura Iancu, Founder of Searchpedia, put it bluntly on LinkedIn: “No more ‘oopsie, we just dropped another core update.’”

Chloe Smith, Strategic SEO Lead at Blue Array, expects pushback: “I expect Google will try to find a way around this.”

Read our full coverage: Google Must Give Notice Before Significant Ranking Changes

—

Theme Of The Week: The Structured-File Ask Keeps Growing

Most of the week circles the same request: publish a structured file for AI to read and host it on your own domain.

llms.txt is the cautionary tale. The file already exists, yet Google says it cannot tell sites apart, and the data shows the bots barely read it. OKF and ARD are that same request arriving fresh, their adoption still unproven. Bing’s new tools sit on the far side of the bargain, measuring whether any of this publishing actually turns into citations.

The request is becoming routine even where the payoff is not. Which of these formats earns its keep is still being decided. Sorting the winners from the files nobody reads is the work this week hands you.

—

Top Stories Of The Week:

• Bing Rolls Out AI Citation Share In Webmaster ToolsMore Resources:Featured Image: PeopleImages/Shutterstock