Industry briefing

November AI search news and what publishers can measure

At a glance — November AI search news changes how publishers can check their visibility. Google added reporting filters and temporary AI Mode threads. Bing described how users can switch AI answers off. Our own citation case covers 21 days, broken into 3 weeks of 7 days. It dates the loss to a week before Google’s core update.

Which November AI search changes should publishers act on?

A Singapore marketing team checking its visibility this month could get three different answers to the same question. Google can show a thread detached from the tester’s account. A Bing user can switch AI answers off. A developer can give Perplexity a URL and have it read the page. Only the first two involve a consumer search. The Bing user may have chosen to see ordinary links alone. A weekly report that gives all three a score out of ten would conceal the most important difference.

We reviewed seven developments announced or documented by Google, Microsoft, OpenAI and Perplexity from 10 to 18 November. We chose changes that affect how a publisher investigates a loss, reproduces an answer or interprets a report. Here is what we would change in a search team’s next working session.

How we selected the announcements

We checked dated releases from the four platform owners against three questions. Could the change move a search result, alter the test environment or change what a report counts? The dates below are the publishers’ announcement dates. Our suggested checks are editorial judgements.

10 November · Date the loss before blaming the Google core update

Google started a global core update covering all languages on 10 November. Its dashboard estimated up to two weeks for the rollout. When we checked on 21 November, it still listed only the start notice. Google Search Status Dashboard.

We tried a simple test on our own timeline. AI Studio’s daily Bing citations fell from 840 on 2 November to 48 on 3 November. That drop happened a week before Google’s announcement, on a different search engine. A report that simply labelled November “core update month” could send an editor to rewrite the wrong pages. Write down the platform, the first affected date and what fell before assigning a cause. Our 21-day Bing record.

11 November · A Bing search can deliberately omit AI answers

Microsoft described a preview extension for Chrome and Edge that lets a user switch Bing’s AI answers on or off. Installing it also sets Bing as the browser’s default search engine. For a single query, adding -ai requests ordinary search results. Microsoft’s Bing announcement.

Before counting a missed citation, record whether there was an AI answer to inspect. Consider ten checks where three show ordinary results only and two of the seven AI answers cite your site. Report both numbers: AI answers appeared in seven of ten checks, and your site appeared in two of those seven. This is a worked example. It shows why an agency should disclose its denominator.

12 and 16 November · Google gave analysts two different ways to trace visibility

On 12 November, Google documented how to identify visits from Google Search using the HTTP Referer header. This gives the team investigating a traffic change a way to recognise referred visits in its logs. An answer that nobody clicks produces no such visit. Google’s dated documentation update and guide to identifying Search traffic.

On 16 November, Google added branded-query classification, web multimodal reporting and multi-country and multi-device filters to the Search Analytics API. Eligible properties can separate brand searches from discovery searches. The webMultimodal type covers Lens, Circle to Search on Android and image uploads in Search. If an ineligible property returns no branded rows, label the segment unavailable. Google Search Central’s API release.

For a Singapore creative studio, the distinction is practical. A buyer who searches the studio’s name after a pitch knows the brand. A buyer who finds a product photograph through Lens has taken a different route. Keep brand demand, visual discovery and referred visits on separate lines. Record the same country, device and dates each time so a new reporting filter does not look like a change in demand.

Google Web query comparison for AI Studio in Singapore, with average position for AI search agency Singapore moving from 3.7 to 15.5
Our Singapore Google Web figures before the API announcement. For “AI search agency Singapore”, impressions rose from 43 to 88 and average position moved from 3.7 to 15.5. There were zero clicks in both 28-day windows, 21 September–18 October and 19 October–15 November. More impressions alone would have given the wrong impression of progress. See the query comparison.

What should a studio check when a buyer searches with a photograph?

A visual-search impression gives an editor a reason to inspect the photograph as well as the landing page. If the buyer is comparing food, check the portion, ingredients and finish against what can actually be ordered. In its Singapore reporting on AI food imagery, The Straits Times examined that gap between an image and the dish. Our practical check is to put the approved product reference next to its published image before counting a rise in visual discovery as a success. The Straits Times’ reporting on AI food images.

16 November · OpenAI changed clinical search in specific products

OpenAI announced changes to how ChatGPT for Clinicians finds, evaluates and cites medical evidence. It also reported clinical-search improvements in ChatGPT for Healthcare. The release names those products. If an agency attributes a change in ordinary ChatGPT recommendations to this announcement, ask it to identify the product used for its observations. OpenAI’s 16 November clinical search release notes.

17 November · Google’s temporary threads make a better comparison possible

Google began rolling out temporary AI Mode threads globally across supported languages on desktop and mobile browsers. Google says the thread is detached from the account, does not draw on past preferences and does not contribute to future personalisation or search history. Google’s temporary threads announcement.

Try this on a Singapore agency shortlist. Enter the same buyer brief in your usual AI Mode session and a new temporary thread. Keep the device, location and wording fixed. Before closing either thread, record each agency named, each page linked and what the answer said about it. Repeat on a second date. One pair can reveal a difference worth investigating. It cannot tell you how often Singapore buyers see that answer.

18 November · Perplexity can read a page you supply

Perplexity’s developer team described its Context API, which accepts up to ten URLs per request and returns material such as Markdown, HTML, links and excerpts. The developer chooses the pages. That makes extraction a useful test for a publisher with a new case study, while source selection in a consumer answer needs a separate observation. Perplexity’s developer announcement.

Give an editor the extracted paragraph without the surrounding page. Can they tell whose result it is, what the number counts and when it was measured? If those details disappear with the caption, put them in the same passage as the figure. You can run that editorial check on the published page before commissioning an API integration.

What did 620 to about 83,000 citations actually tell us?

Our public snapshot shows 620 Bing AI citations for AI Studio’s own website from 24 April to 23 June 2026 and about 83,000 from 23 June to 21 September. Bing displayed the later total as “83K”. These periods cover 61 and 91 calendar days respectively, counting the shared boundary date in each. The rough daily averages were 10 and 912. These are source citations across the Microsoft experiences in the report. Leads and sales require their own records.

Bing citations for AI Studio fell from 840 on 2 November to 48 on 3 November and stayed low for most of the following fortnight
AI Studio’s daily Bing citations, 27 October–16 November 2026. Equal seven-day periods totalled 6,742, 90 and 81. The public case includes the historical screenshot and the daily figures behind this chart.

What would we have wanted on the dashboard before the drop? A measure of dependence on individual pages. In the week ending 2 November, two roadmaps accounted for 5,802 of the 6,474 citations in Bing’s page table, about 90%. The daily overview totalled 6,742 that week. Using the page table’s own denominator reveals how much visibility rested on just two URLs. It gives an editor a reason to broaden the evidence readers can find elsewhere, even during a strong month. Read the full case and inspect the underlying public evidence.

How should a search team compare next week’s results?

Choose the record that can actually answer the question. A named agency, a cited page and a referred visitor can all be different observations for the same brand.

Three decisions and their evidence
QuestionRecord to inspect
Did Google show a page?Impressions for the query, market and search type
Was the business recommended?The saved answer, its wording and the sources it linked
Did a visit produce an enquiry?The referred session and a recorded enquiry event

How to repeat the check

  1. For a lost search position, save the query, country, device, page, dates and replacement result. Open the winning page and write down one question it answers better.
  2. For an AI citation, record the product, session type, exact prompt, presence of an AI answer, named brands and linked source pages. Keep a dated copy that you have permission to retain.
  3. For a business outcome, follow the referring visit to the relevant enquiry event. Mark cases where that link cannot be established. Compare like periods after publication.

For an example of how those distinctions change a buying decision, see our assessment of three different kinds of agency evidence.

About the reporting

AI Studio’s editorial team wrote this briefing. The historical case is credited to Carol Tan, founder of AI Studio. The platform announcements come from the companies named above. The examples concerning our website use figures already published in the Bing case study and Google Web query comparison.

Book an appointment WhatsApp Us