BizBuz

The AI Radar.

What ChatGPT, Perplexity, Claude and Google reward changes constantly. The AI Radar tracks the official sources and the best industry publications, and adapts our scan as soon as the rules shift — with proof attached. That way you know your report follows today's rules, not last year's. Every Saturday we check again; below you'll only find what changed in the past month.

104
signals picked up
104/104
quotes verified
36
sources tracked
14 November 2026
latest sweep
01

How the radar works

Every claim on this page travelled the same road:

  1. 01

    Track the sources

    We follow the official documentation of Google, OpenAI, Perplexity, Anthropic and Bing, plus the best industry sources such as Semrush, Ahrefs, Search Engine Land and Search Engine Journal.

  2. 02

    Literal quote required

    A signal only counts if it comes with a literal quote from the source — no summaries, no interpretations, no rumours.

  3. 03

    Independently verified

    Each quote is then re-fetched independently from the source and compared word for word. If it doesn't match exactly, it doesn't get in.

Only then do we decide: adapt the scan, note a confirmation, keep watching — or deliberately set it aside.

03

New check

New rules from the big players that affect your visibility in AI answers. The scan now checks them.

  • 30 October 2026Source: Google Search Central Blog

    Google plans to support a shared robots.txt signal against AI training from early 2027 (the rule 'User-Agent: __usage_ai-training', still an IETF draft). Mind the trap: if you use that signal to refuse training and don't name Google-Extended separately, Google treats it as blocking Google-Extended. You then also drop out of Gemini app answers that draw on your site. Google Search and your ranking are not affected.

    “When a website uses this policy token to opt out of training in its robots.txt file (and Google-Extended is not explicitly specified), Google plans to interpret that as if the publisher had disallowed Google-Extended.”

    Literal quote — Google Search Central Blog

    What we did with it: New in the scan since 1 November: we recognise the training signal and warn when your robots.txt blocks Google-Extended, directly or through this signal without a separate rule. It stays a tip without points: not wanting AI training is your choice, but you see in black and white that you give up the Gemini app for it.

  • From February 1, 2027 (on each new crawl) Microsoft will read two new preferences. You refuse AI training via robots.txt or per page. Whether your pages are used in Bing and Copilot AI answers (grounding) is set per page only, with a Bingbot meta tag or header ('ai-grounding:disallow' or 'disallowextended'). Both default to allow. Refusing grounding significantly reduces your visibility in AI answers, says Microsoft, without affecting regular Bing results. The old NoArchive and NoCache tags no longer count for training or grounding after March 1, 2027.

    “If you select DisallowExtended or Disallow for grounding, you will see reduced brand visibility in AI experiences.”

    Literal quote — Bing Search Blog — training en grounding

    What we did with it: New in the scan since 1 November: we look on your homepage and in server headers for 'ai-grounding:disallow' or 'disallowextended' for Bingbot, and for old NoArchive/NoCache tags best replaced before March 2027. Most small businesses need to do nothing: without a tag you're on allow.

04

Confirms our approach

Recent publications that back up what the scan already checks.

  • A US study (480 questions, 40 cities, 12 local sectors, about 2,900 answers from ChatGPT, Gemini, Perplexity and Google, one week in November) found that businesses with more Google reviews, a higher rating and a recent review were recommended more often. Having your own website went together with recommendations by Gemini (about 2.25 times the odds) and Google's AI Overviews, but not ChatGPT. These are correlations, not proven causes; the author sells review-reply software, and the thresholds (such as 250 reviews) don't carry over to Belgium.

    “It mattered for Gemini (about 2.25 times the odds) and for Google's AI Overviews, but made no measurable difference for ChatGPT once other things were equal.”

    Literal quote — Revfer — wat lokale zaken aanbevolen krijgt door AI

    What we did with it: Nothing changed in the scan. It confirms why we check both your Google reviews (count and rating) and your own website: each assistant leans on different signals. Candidate check (V1): checking how old your latest Google review is. Not in the scan yet.

  • In its quarterly results of November 5, Tripadvisor wrote off $369.5 million on its 'Hotels and Other' segment, the classic hotel and review business. The reason, according to the company itself: fewer and fewer visitors due to 'changes in the search landscape'. Revenue fell 11%, hotel revenue 23%. Tripadvisor does not name AI as the cause; these are company-wide figures, not specific to Belgium. In the same release: the sale of TheFork to American Express closed on November 1.

    “The impairment was primarily driven by a sustained decline in financial performance and reduction in forecasted profitability, largely due to the multi-period acceleration of traffic headwinds resulting from changes in the search landscape.”

    Literal quote — Tripadvisor — resultaten Q3 2026 (SEC)

    What we did with it: Nothing changed in the scan. It confirms why we check your own site and your Google Business Profile, not just your listing on a platform: if even Tripadvisor is losing visitors as search changes, a spot on such a platform alone is not enough.

05

On the radar

Developments we track before acting on them. Numbers that matter.

  • On November 10 OpenAI launched 'Businesses in ChatGPT' (beta). Through Business Hub you claim your business profile in ChatGPT; after identity verification you manage your details, connect a product feed and see how often you're mentioned and how many clicks to your site that brings. But: local stores and restaurants can for now only claim an existing US listing, so a Belgian local business can't use it yet. Belgian online shops or online services may be eligible. According to OpenAI, claiming doesn't guarantee extra visibility.

    “Local stores and restaurants can currently claim only an existing U.S. listing.”

    Literal quote — OpenAI Help Center — je zaak in ChatGPT

    What we did with it: Nothing changed in the scan. Until you can claim your business in ChatGPT, ChatGPT takes its info from what's on the web: that's why we check whether your own site states your details clearly and readably. If claiming comes to Belgium, we'll report it here.

  • On November 10 Google started the 'November 2026 core update', a broad change to rankings in regular search results, worldwide and in all languages, so also for Dutch- and French-language sites in Belgium. The rollout can take up to two weeks and was still ongoing on November 14. Google says nothing about an effect on AI Overviews.

    “which applies globally and to all languages. The rollout may take up to 2 weeks to complete.”

    Literal quote — Google Search Status Dashboard — November 2026 core update

    What we did with it: Nothing changed in the scan. If your position in Google fluctuates in the coming weeks, this is a likely reason; ideally rescan once the rollout has finished.

  • Google says its new Gemini 4 Hydrogen model has been running in Google Search 'globally for all users' since November 12. According to Google it asks more sub-questions at once and so shows a broader range of links, including to sites you might not otherwise have found. Google doesn't say which AI feature this sits in exactly, nor what it means for Belgium or for Dutch and French; 'more different sites' is Google's own claim.

    “By more thoroughly understanding your question and conducting simultaneous queries across all the relevant subtopics, Search helps you find sites you previously might not have discovered.”

    Literal quote — Google — Gemini 4 Hydrogen in Search

    What we did with it: Nothing changed in the scan. It confirms why we check whether every service you offer is on your site as clear text: the more sub-questions Google asks, the more likely one of them is about exactly your service in your town.

  • Since November 9, Google's blocks with comparison sites (aggregator unit) and with direct providers (supplier unit) also show results for searches about activities and outings ('things to do'). This applies to users in the European Economic Area, so also in Belgium; local businesses were already included since September. An attraction or activity provider can appear as a direct provider without a separate feed: Google uses what it can crawl from your site. The provider block only appears when the comparison block does, and Google doesn't say which signals decide who gets in.

    “The supplier unit is available to users in the EEA when the aggregator unit appears for queries related to hotels, flights, long distance trains or buses, products, local businesses, and things to do.”

    Literal quote — Google Search Central — supplier unit

    What we did with it: No new check. For attractions, tours and activities: we already check whether Google can read your site and whether your offer, prices and booking option are on your own site as text, which is exactly what Google draws on here.

  • On November 11 the European Commission proposed the Digital Fairness Act: a regulation that replaces the current Unfair Commercial Practices Directive and applies to every business selling to consumers, not just platforms. New elements include a ban on misleading design tricks (dark patterns) and rules on prices, influencers and subscriptions. The rules on reviews are not new: writing or commissioning fake reviews, showing only positive ones and deleting negative ones has been banned since 2022, and anyone displaying reviews already has to say whether and how they check that they come from real customers. It is only a proposal; Parliament and Council still have to approve it.

    “the DSA applies only to online platforms, while the DFA would apply to all companies engaging in business-to-consumer commercial practices”

    Literal quote — Europese Commissie — Q&A Digital Fairness Act

    What we did with it: Nothing changed in the scan. Nothing changes for you for now: collect reviews from real customers, don't buy any and don't delete negative ones. Genuine, recent reviews also count with AI assistants.

  • Since November 4, Anthropic explains how its cloud-hosted browser agent identifies itself (used by Claude Duet and other Claude services, among others, to complete tasks on websites). It uses a regular Chrome browser with the extra token 'Claude-Agent' and comes from IP addresses Anthropic publishes. That traffic is separate from ClaudeBot, Claude-User and Claude-SearchBot and is not controlled through robots.txt, but through your firewall or bot protection. If you block it, Claude users can't get anything done on your site through such an agent. Claude Duet itself is currently only available in the US.

    “Allowing or disallowing this traffic will control whether your users can ask their Claude services that leverage this type of agentic browser use to complete tasks on your website.”

    Literal quote — Claude Help Center — Claude-Agent

    What we did with it: Nothing changed in the scan: we check in your robots.txt whether Claude's search bots are allowed, but we can't see firewall or bot-protection rules. Candidate check (V2): checking whether your site blocks 'Claude-Agent' traffic. Not in the scan yet.

  • 30 October 2026Source: OpenAI — AI training preferences

    The same day OpenAI also announced it will honour that shared robots.txt signal against AI training, 'in the coming months'; nothing is live yet. The existing per-bot rules still apply. According to the IETF draft the signal is not a crawling instruction, so using it doesn't remove you from regular search results.

    “We aim to implement the signal in the coming months, and will share another update with details about our implementation when we do so. This new signal will not replace the controls we already provide through robots.txt.”

    Literal quote — OpenAI — AI training preferences

    What we did with it: Nothing changed in the scan: we keep checking ChatGPT's search bot separately. Once the signal is live, it's a clean way to say 'no training' without accidentally blocking the search bot.

Is your site ready for today's rules?

The scan checks your site against the current state of this radar — in a few minutes.