Bing names the AI queries where rivals get cited, not you

Bing Webmaster Tools now lists the AI queries where you are not cited

Why it matters for your score

Topics Portfolio, a beta tab in the AI Performance report, groups your grounding queries into AI-classified topics and shows topic size, how many sites get cited, and a Top Opportunity Queries list of questions where other sites are cited and you are not. Citation volumes show as ranges, date windows run 1 to 3 months, and the tab only appears for sites with enough data. This is a first-party view of your AI citation gap at no cost, so verify your site in Bing Webmaster Tools this week even if you wrote Bing off for traffic years ago.

Amex publishes a framework for proving what an agent was told to buy

Why it matters for your score

A Create Intent API records the customer's instructions and limits, including merchant, maximum budget and preferences. The customer then authenticates and approves them, and the system issues an Intent ID and a Proof of Intent Token that is exchanged for a payment credential the agent uses at checkout instead of a card number. Amex says it wants this to become an industry standard and names EMVCo and the x402 Foundation as venues. Agent purchases are heading toward constraints that are signed before checkout, which means your prices, availability and product facts have to be correct at the moment an agent reads them, not at the moment a human phones you.

Google confirms PDFs are converted to HTML and can surface in AI Mode

Why it matters for your score

On the 13 October episode of Google's own Search Off the Record, Martin Splitt and John Mueller said Google renders a PDF, converts it to HTML before indexing, and then treats it much like any other document, so converted PDFs can appear in AI Overviews and AI Mode. Googlebot fetches the first 64MB of a PDF, neither host could name that documented limit on air, and Google refetches PDFs rarely because it assumes the file will not change. Mueller suggested publishing an updated PDF under a new filename. If your prices, specs and policies live in PDFs, they are answerable content, and Google may be answering from the version you replaced months ago.

Gemini 4 Argon is now answering AI Mode queries in Search

Why it matters for your score

Google made Argon available in AI Mode in Search and in the Gemini app for AI Pro and Ultra subscribers, with API access for developers, and pitches it at multi-step tasks and instruction following. The model deciding which 3 companies get named in your category changed this week. Any shortlist you captured in September describes a system that is no longer running, so rerun your own category questions before you quote that data to anyone.

Google adds gluten-free, dairy-free, halal and kosher attributes to Business Profiles

Why it matters for your score

Restaurants can now set 4 more dietary attributes alongside the vegetarian and vegan ones that already existed. Attributes are machine-readable facts rather than marketing copy, and a dietary constraint is exactly the filter an assistant applies before it names anywhere at all. Fill in every attribute that applies to you, because a blank field reads as a no.

Citation gap reporting is turning into a platform feature rather than a product category.

The engines themselves are starting to tell you which questions you lose, which moves the hard part from measuring the gap to closing it.

Agent commerce is standardising on proof of instruction rather than proof of identity.

The question moving through payments is no longer who is buying, it is what the buyer actually authorised.

Formats you treat as archives are being read as current fact.

PDFs, profile attributes and product feeds all get parsed as the truth about you, and nobody refetches them on your publishing schedule.

The model behind the answer now changes faster than most reporting cycles.

An AI visibility snapshot older than a month describes a system that has already been replaced.