Anything Engine Optimization The rolling record of the AI-search industry.
Send a tip
A single summation sign on a deep colour field — the house card for the weekly briefing.
AEO Wire
AEO Wire · Nov 2 AEOGoogle SearchResearch

Last Week in AI Search: November 2

What this answers

How many items did the AI-search wire publish this week?
46 items between 2026-10-26 and 2026-11-01, across the lanes AEO, Google Search, Research. Each is dated and sourced on the wire.

In this issue: Why SEO reporting can't keep up, and six questions to ask your tracker · Plus: OpenAI, Anthropic, Google and Microsoft back one AI-training opt-out

The Story

50% of AI citations vanish overnight. How to keep your AI visibility tracker honest.

What can the weather in Cleveland, Ohio, teach us about AI search?

I’m sorry, that might be the most LinkedIn sentence I’ve ever written, but I can’t not make the connection.

“Don’t like the weather in Cleveland? Just wait 5 minutes; it’ll change.”

Weirdly, that’s how AI citations are currently behaving. Don’t like that your brand wasn’t cited? Check again tomorrow and roll the dice. Love that you’re cited AND linked? Check again tomorrow for a real gut punch.

I keep hearing from people who have SEO in their titles that AI search is SEO, but the data doesn’t back that up. AI search needs studies, not snapshots. It needs an average, not a rank. It’s dynamic, fluid, and honestly quite frustrating.

Ahrefs asked five AI tools the same questions daily for 20 days. They tracked every website each tool cited on the first day. By day two, about half of those citations were gone. Google’s AI Overviews dropped 50% of them, AI Mode 58% and ChatGPT 59%.

Most of them eventually came back. About three out of four citations that vanished came back in 20 days. Half returned within just three days. Ahrefs didn’t change a thing between checks.

Sistrix, after tracking 82,619 prompts for 17 weeks, put it plainly: a single citation placement “is not a reproducible result but merely a snapshot.”

A single snapshot doesn’t tell you where you stand. It tells you where you stood once.

That doesn’t mean it’s all noise. Volume is how you sift it. Sistrix estimated what that does to a score. Take a brand that truly appears in 25% of answers. Ask one prompt every day for 20 days, and in about two of three cases the score lands somewhere between 0% and 59%. That’s a range running from invisible to more than double the brand’s real rate. Ask 100 different prompts, and it tightens to 21% to 29%.

Breadth tells you how widely you appear. Repetition tells you whether that presence holds. In Ahrefs’ data, a citation that lasted one day and then dropped stayed gone 34% to 41% of the time. One that lasted eight to 14 days almost always came back. In Google’s AI Mode, Sistrix found 86% of prompts kept a stable core of a few websites while the rest rotated. But repetition can’t rescue a narrow set of prompts. It only tells you how that narrow set behaves. The signal is there. One look just won’t show it to you.

How AI search breaks SEO’s three rules

SEO reporting was built on three things that mostly held true. AI search breaks all three.

  • The results held still, mostly. Rankings moved, but a rank tracker could check a page of ten links once a day and report a position. An AI answer is written fresh every time. Seer Interactive ran one prompt twice, eight minutes apart, and the two runs shared no searches.
  • You knew most of what people searched. Search Console reports the real queries people typed. Even when Google hid search terms from analytics in 2011, labeling them “(not provided),” SEOs still had that data. ChatGPT has no Search Console, so AI-visibility tools write their own prompts.
  • Most visits arrived labeled. Someone clicked, and analytics could usually say where they came from. In October, ChatGPT stopped tagging its links and stopped passing a referrer. Adobe saw ChatGPT-attributed visits fall by about 31,000 a day across roughly 500 pages, while Direct rose by about 26,000.

So an SEO report starts from a count of real searches and asks what it means. An AI-visibility score starts from a set of prompts someone designed, and it describes that set. That’s a very different kind of number.

Six questions to keep your tracker honest

Sooner or later, someone will show you an AI visibility score or a before-and-after. Ask these first.

  • How many prompts? One prompt erred by ±34 points in Sistrix’s data, and 100 prompts by ±4.2.
  • Who picked them? Three times out of four, a brand looked better against its rivals in a prompt set it chose itself than in Sistrix’s market-wide sample. The prompts you choose can flatter you.
  • How many runs, before and after? Citations can move overnight on their own. A single check on each side can’t tell your change apart from that churn. Sistrix recommends comparing weekly averages.
  • Does it count each engine separately? In Sistrix’s 17-week study, only 17% of the websites Google’s AI Overviews and AI Mode cited for the same prompt showed up in both. One blended number hides that.
  • Did it actually ask an AI? Some “visibility scores” grade your robots.txt, sitemap and schema. That’s an SEO audit. It can tell you whether a crawler can get in, not whether an AI names you.
  • What else could explain it? Repeated measurement can show a change is real. Comparing against something you left alone makes the cause far more believable. SEO learned that one the hard way, too.

The Follow-Up

ChatGPT’s rewritten searches add brands, and your memory

Last week’s Story found ChatGPT never searched a UK local question word for word. Two studies this week show more of what goes into the rewrite.

Graphite, an agency that sells AEO services, counted about 957,000 of ChatGPT’s background searches from 22,953 comparison prompts that named no brand. 59.6% of those searches named one anyway, and 22.7% searched inside a single site.

Profound gave ChatGPT synthetic users with saved memories. When a memory was relevant, 19% of ChatGPT’s searches referenced it. Two answers for the same user overlapped by about 42%. A user with relevant memory and one with none overlapped about half as much. (Profound’s users and their memories were synthetic.)

The Numbers

  • 56.7% of the statements Google’s AI Overviews cited were supported by the page they cited, across 3,611 health questions. Overviews appeared on 96.4% of them. (arXiv preprint, not peer reviewed) Read →
  • U.S. organic search visits fell 13.2% between August 2024 and August 2026 across 8,658 sites. AI referrals grew fastest of any channel and still reached only about 0.5% of all visits. (Growth Memo analysis of Similarweb estimates) Read →
  • Reddit’s U.S. logged-out daily visitors fell from 30.1 million to 27.3 million in one quarter. Its CEO said Google’s AI Mode “does not drive traffic.” (Reddit Q3 call) Read →

The Rest

All 46 items from October 26–November 1, dated and sourced: the week on the wire · State of Search · Corrections · Archive. Seen something the record should hold? Reply to this email; tips are credited by name.

The full wire

47 items published October 26–November 1. Every item dated, sourced and linked. An item can carry more than one lane, so lane totals exceed the count.

Google Search · 16

Research · 12

Law & Money · 7

Engines · 7

Tools & Vendors · 4

GEO · 1

Read the complete archive →

Why it matters: The one place on the wire where the week is read rather than reported — compiled from the published record, framed by the editor, checkable line by line.

Glossary: Answer engineAEOGEOSEOAI crawlersRAGAI visibilityAI OverviewsAI ModeGroundingGoogle Discover

Via AEO Wire ↗

Posted to the wire November 2, 2026. Edited by Joe Balewski.

More on the wire

← Back to the wire