AI search this week: citation gaps, official pages, and crawlers that get graded

The question

What changed in AI search in mid-October 2026, and what should a small business do about it?

The short answer

Bing Webmaster Tools added a Topics Portfolio beta that shows topics and "opportunity queries" where other sites are cited in AI answers and yours is not. Google is testing a section at the top of AI answers for the one or two official pages for a task. Cloudflare and OpenAI reported a pilot that brought previously unknown pages into OpenAI's search index, and Cloudflare is grading verified bots internally on whether they respect robots.txt and slow-down requests, with public scores planned. Check your Bing opportunity queries, make your task pages unmistakably official, keep your site reachable for search crawlers, and decide your bot policy on purpose.

TL;DR: Four changes this week, all about which pages AI answers can find and show. Bing added a Topics Portfolio beta that lists AI-answer queries where other sites are cited and yours isn't. Google is testing a top section for the one or two official pages for a task. Cloudflare and OpenAI say a pilot brought previously unknown pages into OpenAI's search index. And Cloudflare is grading verified bots internally on whether they respect site preferences, with public scores planned. Our read: open Bing's report, make your task pages obviously official, and set your crawler rules on purpose.

The default: someone asks "are we showing up in AI answers?", a person types three questions into ChatGPT, and the screenshot goes in the deck.

That was never a measurement. This week, two engines and a network gave you better tools than a screenshot.

What changed

Bing shows where other sites get cited and you don't

Bing Webmaster Tools already has an AI Performance report: the grounding queries where your site appears in AI answers, and the pages cited. On 13 October Bing added a Topics Portfolio tab, in beta.

It groups those queries into topics. For each topic it shows:

• Topic Size: approximate total citations across all sites, so a rough measure of demand.
• Sites Cited: approximately how many sites appear, so how crowded it is.
• Queries: approximately how many distinct grounding queries it holds.

Below that sits Top Opportunity Queries: queries where other sites appear and yours does not, with citation volumes shown as ranges. Each topic page splits what you already have from what is still open. Bing's Getting started notes let you pick a one-, two- or three-month window for the report.

Bing is blunt about the limits. Topics are sorted by AI, so a query can land in the wrong one. Numbers are approximate and meant as "directional signals for planning, not as a precise ledger." Branded queries for a competitor can show up as "opportunities." And topics without enough data are hidden, so smaller sites may see a short list or no tab at all.

Google is testing an "official links" section in AI answers

On 15 October Google announced changes to AI Mode and AI Overviews:

• More inline links, shown higher in answers more often.
• A test of a dedicated section at the top of an answer for the "one or two official places on the web to complete a specific task," with context on why each link is useful.
• Carousels for timely articles rolling out to more countries and languages. They highlight Preferred Sources, a feature Google says has grown more than 7x in users since the start of the year (no starting or absolute count given).
• More prominent related-article links at the end of some answers, with image thumbnails.
• A test of clickable creator and publisher names on video results in AI Mode, linking to Search profiles.

Google gave no click or traffic numbers.

OpenAI's crawler gets a new way to find pages

On 15 October Cloudflare and OpenAI shared early results from a research pilot that started in July. Signals from Cloudflare's network help OpenAI's search crawler discover pages. OpenAI then evaluates and ranks those pages on its own.

The pilot used about 1% of Cloudflare's network and, they say, brought hundreds of millions of new pages into OpenAI's search index. Their examples are the pages few sites link to: a manual for one washing machine model, current passport fees, a restaurant's latest menu.

They also describe what makes a URL look like noise: bots requesting files that don't exist, sites that redirect almost any address to the homepage, and addresses that lead to copies of the same content. They say a signal about a page "doesn't grant permission to access it," and that the research uses metadata about the web for search and discovery, rather than sharing page content for model training.

No publisher traffic or citation numbers yet.

Cloudflare is grading bots on manners

On 16 October Cloudflare introduced a "trust band" for Verified bots. Being verified tells you who runs a bot. The trust band tracks what it does over time. The initial signals are whether the bot respects instructions and whether it responds to direct feedback. In practice that means consulting robots.txt, and slowing down or pausing after an HTTP 429 "Too Many Requests" response.

Cloudflare says it already uses this internally, plans to publish scores on Radar, and plans to let customers choose a trust threshold for how verified bot traffic is handled. It gives no date. It also notes that it removed Perplexity's bot from the Verified list in 2025 after the bot sought access on sites that opted out of scraping and then tried to get around the blocks.

What it means for a business

This section is our read, not the sources'.

Visibility in AI answers is becoming something engines report back to you. Bing's tab is a first-party view that says "here are the questions in your area where you're not cited." That beats any screenshot. It is also a beta, sorted by AI, with ranges instead of counts. Treat it like a lead list, not a scoreboard.

Some questions have an official answer, and Google is testing giving that answer its own box. If you are the official source for something (your booking page, your returns policy, your price list, your service area), the page should say so plainly. A page that has to be scrolled to figure out whose it is probably won't be picked as "official." That last part is our inference; Google didn't say how it picks.

Backlinks are not the only discovery path the pilot describes. It is about pages with few links, which is exactly what most small-business sites are made of. Whether you are in that 1% cohort is not something the post says how to check. What you control is whether a crawler that arrives finds a real page or a redirect to your homepage.

Your bot policy is turning into a public decision. If Cloudflare publishes trust scores, "block everything" and "allow everything" both become choices someone can see the effects of. Decide which crawlers you want answering questions about you, on purpose.

Nobody this week promised more traffic. Please hold the champagne.

What to do this week

If you only do three: 1, 3, and 5.

1. Open Bing's AI Performance report. Sign in to Bing Webmaster Tools, open AI Performance, and look for the Topics tab. If you have it, write down your top five opportunity queries. Cross out opportunity queries that are a competitor's brand name. If you don't have the tab, that may mean Bing doesn't yet have enough site presence or ecosystem activity to show it. Note it and check again next month.
2. Turn opportunity queries into page work, not new posts by default. For each query, ask whether an existing page already answers it. If yes, put the answer in the first two sentences of that page. If no, that is your next article.
3. Make your task pages unmistakably official. List the three to five tasks people come to you for: book, price, return, apply, find hours. Give each its own page with your business name and the task in the title, the answer near the top, and current dates or prices. One page per task, one task per page.
4. Clean up the noise crawlers trip on. Old URLs should return a real 404 or redirect to the matching page, not dump everyone on the homepage. Remove duplicate copies of the same page, or point them at one version with a canonical tag.
5. Check that search and AI crawlers can actually reach you. Read your robots.txt. If you are on Cloudflare, check what your bot settings block. A site that returns 403 to the crawler you want is invisible to that answer engine, however good the page is.
6. Decide your crawler policy and write it down. Search crawlers that send people to you, AI training crawlers, and AI agents acting for a user are different jobs. Allow or block each on purpose, and revisit when Cloudflare publishes its trust scores.
7. Retire the screenshot. If you report AI visibility, report the Bing topics and opportunity queries with the date range, plus Search Console for Google. Say when a number is a range.