Editor's note
Search was the main thread today. Google put its new Gemini 4 Argon model into AI Mode, and Bing added a report that shows a site's AI-search visibility by topic. Alongside that came practical, low-key updates — new dietary attributes in Google Business Profile and Google's own explanation of how PDFs are indexed. Getting ready for AI search still starts with keeping basic information accurate and well structured.
AIGoogle Blog
Google launches Gemini 4 Argon, rolling it into AI Mode in Search and the API
On October 13, Google announced Gemini 4 Argon, a model aimed at complex, multi-step tasks. It is available in the Gemini app and AI Mode in Search (for AI Pro and Ultra subscribers), the paid tier of the Gemini API, and AI Studio. Introductory API pricing is $2 per million input tokens and $10 per million output tokens, rising to $4/$20 on November 30.
Context
Google had recently brought its Gemini 3.8 models into AI Mode, but the Flash version drew complaints for omitting links in answers. Argon is positioned for coding, professional work in fields like law and finance, and agent-style autonomous tasks, and reports say its AI Mode answers do include links. Putting a top-tier model directly into Search could change both answer quality and how sources get cited.
HaLVision's take
For small-business sites, the key detail is that AI Mode answers carry source links. Which pages get picked as sources tends to depend on clarity and structure, so the practical move is to keep pages that state services, prices and location in concrete terms. Anyone budgeting for API use should note that the introductory price doubles at the end of November.
WebBing Webmaster Blog
Bing Webmaster Tools adds Topics Portfolio beta to show AI-search visibility by topic
On October 13, Microsoft Bing added Topics Portfolio, in beta, to the AI Performance report in Bing Webmaster Tools. It groups the queries where a site was used to ground AI answers into thematic clusters, showing each topic's size and how many sites compete in it. It also lists opportunity queries where competitors are cited but your site is not.
Context
In AI search, rankings and clicks alone don't show which subject areas a site actually has presence in. Bing already offered a report on citations in AI answers, and this update rolls that data up by theme. Bing notes that the topic classification is AI-generated and can be imperfect, and that smaller sites may not have enough data to be included.
HaLVision's take
This is one of the few first-party ways to see which topics your site gets cited for in AI answers. Bing's share is modest in many markets, but the data still hints at how Copilot-style answers draw on your content. The opportunity queries look directly useful for deciding which pages to expand or which articles to write next.
WebSearch Engine Roundtable
Google Business Profile expands restaurant dietary attributes, adding gluten-free, halal and more
Google is adding gluten-free, dairy-free, halal and kosher dietary attributes for restaurants in Business Profiles. They join the existing vegetarian and vegan options. Owners can tag individual dishes through the menu editor.
Context
Google's Lisa Landsman announced the change on LinkedIn, and Search Engine Roundtable reported it on October 13. Tagged dishes can surface for searches like ‘gluten-free dinner near me’ on Maps and Search. The report notes that claims such as kosher rely on the restaurant's own declaration, with little third-party verification.
HaLVision's take
From a local SEO standpoint, this is a practical update owners can act on right away. Restaurants with qualifying dishes should make sure their menu is entered in the profile and tag those items, which can open up more filtered searches. Because it is self-reported, accuracy matters — a wrong tag costs more trust than it gains.
WebSearch Engine Roundtable
Google explains how Search handles PDFs on its podcast: default to HTML
Google's Martin Splitt and John Mueller discussed PDF indexing on the Search Off the Record podcast. PDFs are converted into an HTML-like representation; text links inside them are crawled, but embedded images can't be indexed for Image Search. PDFs are recrawled less often than HTML and can appear in AI Overviews.
Context
Many sites still publish brochures and price lists only as PDFs, so how Google treats them is a common SEO question. The episode also covered practical drawbacks: scanned PDFs can't be read without OCR, and mobile users must download the whole file before viewing. Google's bottom line is to use HTML by default and reserve PDFs for forms, print layouts or offline documents.
HaLVision's take
For clients who post price lists or service guides only as PDFs, we'd recommend putting the key information on an HTML page as well. The tip to change the filename when publishing an updated PDF is useful, given the slower recrawl. And it's safest to assume a scanned PDF is effectively invisible to both search and AI.
AIMicrosoft
Microsoft unveils Maia 300 AI chip, plans multi-gigawatt deployment from 2027
On October 12, Microsoft announced its in-house Maia 300 AI chip and plans to scale Azure Maia to multi-gigawatt deployments starting in 2027. The company says Maia 300 delivers about 40% more tokens per megawatt than leading GPU systems. The current Maia 200 already serves Copilot, GitHub Copilot and Microsoft Foundry.
Context
Major cloud providers are building their own chips to cut the cost and power draw of AI inference and reduce reliance on outside GPU vendors such as Nvidia. Maia 300 is inference-first, designed for long-context and memory-heavy workloads, and is slated to become the main inference platform for Microsoft's products. Full deployment starts next year, and the performance figures are Microsoft's own.
HaLVision's take
Small businesses won't pick chips themselves, but cheaper inference on the cloud side could eventually show up in Copilot and API pricing or usage limits. There's nothing to change today — this is useful background for reading AI service price changes over the next year or so.
AIGOV.UK
UK steps up national defences against AI risks, including incidents involving autonomous agents
On October 12, the UK government announced measures to strengthen defences against AI-enabled cyber threats. It will strengthen the Cyber Security and Resilience Bill, including how serious incidents and near-misses are identified and reported. It also said the AI Security Institute is evaluating the latest frontier models with developers and the US government.
Context
As AI agents carry out multi-step tasks on their own, there is concern that incidents could escalate faster than traditional cyber attacks. The UK says pilots using frontier AI for defence have already found and fixed more than 1,000 vulnerabilities in government systems. On the same day, Microsoft President Brad Smith also called for tougher AI safety rules, a sign the regulatory debate is getting more concrete.
HaLVision's take
As more businesses hand work to AI agents, expectations around incident reporting and shutdown procedures are likely to spread to other countries too. Even small firms can start now by giving AI tools only the permissions they need and keeping a record of what each tool is allowed to do.