Anything Engine Optimization The rolling record of the AI-search industry.

Corrections

Every correction the wire has run, in one place. Corrections are appended to the item and dated — never silently edited. This log exists whether it's empty or not.

SparkToro and Similarweb: open-web traffic flat since 2021 as web users grew 24%

  • 2026-11-12: This item said total visits to the top million sites stayed about flat from 2021 to 2026 and that the post's text did not report a decline in visits. SparkToro's post, as it reads on November 12, is headlined 'Traffic to the Open Web Has Fallen 12% in 5 Years' and reports that visits peaked in Q1 2022 at 2.243 trillion and were 12.2% below that peak in Q3 2026 (1.970 trillion; low point Q1 2026, down 14.7%). It also gives 29.2% as the fall in visits per user, from about 24% more estimated web users (DataReportal; 4.95 billion to 6.15 billion). We have not established whether the post was revised after November 10. The headline above and the 'flat' wording are superseded by SparkToro's stated figures; the browser-only, global-totals and unknown-cause limits stand.

Zyppy opt-out test estimates AI Overviews and AI Mode drove 21% of one site's clicks

  • 2026-11-14: Merged: the wire first reported this test on October 30, from Search Engine Roundtable's pickup of Shepard's LinkedIn post, with his estimate of about 0.9% click-through on AI impressions. That was the same experiment as this item, so the October 30 item now redirects here.

People Inc. core sessions fall 26% in Q3, citing AI search and AI Overviews

  • 2026-11-07: The reusable social copy described Google as 21% of traffic. People Inc.’s Q2 presentation reports Google Search as 21% of Core Sessions, not all traffic. The social field now preserves that scope; the live social posts require a visible correction reply.

OpenAI and Anthropic join Google, Microsoft in backing one robots.txt AI-training opt-out

  • 2026-11-05: The wire covered this announcement twice. The Oct. 30 item, "Google plans to support shared AIPREF AI training opt-out in robots.txt," now redirects here, and its details are incorporated.

Digital Content Next brief puts publisher traffic decline at 22% in 2026, OpenAI court told

  • 2026-11-05: The wire covered these amicus briefs in separate items. "News/Media Alliance files amicus brief in publishers' OpenAI, Microsoft case" and "Open Markets Institute files amicus brief backing publishers against OpenAI, Microsoft" now redirect here, and their details are incorporated.

Alphabet Q3: Google Search revenue growth slows to 12% as AI Mode ads go global

  • 2026-11-05: The wire covered this earnings call in three items. The Oct. 29 items, "Google says lower AI Mode monetization weighed on Search revenue growth" and "Pichai says Google is bringing agentic features to Search next, with monetization after," now redirect here, and their details are incorporated.

Google opens Connected Apps Platform so developers can add apps to Gemini and AI Mode

  • 2026-11-05: The wire covered this launch twice. The Oct. 31 item, "Google's Connected Apps policy bars deceptive descriptions to win AI Mode selection," now redirects here, and its details are incorporated.

Last Week in AI Search: October 26

  • 2026-10-31: The archived briefing gave conflicting totals of 52 and 55 items for October 19–25. Its Q&A, footer and full-wire list now agree on the 54 distinct news items in the current archive after the duplicate Google robots.txt story was merged. The original draft metadata has been reconciled to that count.
  • 2026-11-07: A subsequent archive merge combined the duplicate Meta scraping-terms reports. The current October 19–25 archive therefore contains 53 distinct news items, rather than the 54 recorded in our October 31 correction. The Q&A, full-wire list, footer and draft metadata now match that count.

Publishers ask EU, UK to halt Meta AI-scraping terms

  • 2026-11-05: The wire covered this pushback in separate items. The Oct. 22 item, "News/Media Alliance demands Meta withdraw AI scraping terms," now redirects here, and its details are incorporated.

ChatGPT cites official sources over local news, study finds

  • 2026-10-26: This item first said ChatGPT cited local news sources in 0.9% of answers across 30,000 prompts. The 30,000 prompts went to five AI assistants; ChatGPT received 6,000 of them. Press Gazette describes the 0.9% as the share of questions where ChatGPT went to local news, not a share of cited answers, and does not define it further; Searchable's own data page does not publish the figure. The item, its Q&A and the /numbers/ entry now say so. The item also notes that Press Gazette doesn't say which assistants the 18.8% and 21.2% citation shares cover.

Google: robots.txt limits what gets crawled, not what AI does with the data

  • 2026-10-31: The wire covered the same October 21 Google Search Central video twice. The October 23 copy now redirects here; its additional details and Search Engine Journal attribution have been incorporated. This is a historical account, not a statement of current training-control standards.

ChatGPT Ads will auto-enable creative enhancements Oct. 30

  • 2026-11-05: The wire covered this rollout twice. The Oct. 30 item, "ChatGPT Ads creative enhancements switch on by default for all campaigns," now redirects here, and its details are incorporated.

Analysis: GA4 Direct traffic shows little hidden AI activity

  • 2026-10-23: The wire published this analysis twice, here and on Oct. 21 as "Analyst finds no reliable sign of AI traffic hiding in Direct"; that copy now redirects here. It said engaged Direct traffic "rose just 2.3%" where this item says it fell 1.0%. Both are Indig's figures, measured two ways: engaged Direct changed −1.0% in monthly counts and +2.3% in average daily volume, normalized for days per month. That copy also reported that at one site GA4 recorded 409,000 Direct sessions, of which only 198 counted as engaged.

DataDome: Meta's crawlers drove half of Q3 AI agent traffic

  • 2026-10-23: The wire published this report twice on Oct. 19; the other copy, "DataDome: Meta's crawlers made 49.7% of Q3 AI agent traffic", now redirects here. Its figures matched this item's. It also reported that in a sampled subset of Muse's traffic, 99.3% came from cloud or CDN infrastructure rather than end-user devices, per DataDome.

Sites with OpenAI deals let ChatGPT's search in and keep Claude's out

  • 2026-10-19: This item's summary first said which assistant someone uses "now decides which publishers it can read." robots.txt rules govern a company's own crawlers, not every route to a publisher's content: licensed partners often receive feeds or APIs, and AI search tools can draw headlines and snippets from other search indexes. The summary now says what the data shows, and the item adds the training-bot and Google-Extended counts and the check that the Claude blocks are named, not blanket.

Chartbeat: Google traffic to publishers fell 40% in a year

  • 2026-10-17: This item originally said small publishers lost about 60% of their search referral traffic, medium publishers 47% and large publishers 22%, and attributed that breakdown to Chartbeat's 2026 Publisher Playbook. The playbook gives no breakdown by publisher size. The 60% figure comes from earlier Chartbeat data that Axios reported in March, and Search Engine Journal says the playbook doesn't break down its data by size. The breakdown has been removed from the item, its Q&A and its summary.

Ahrefs: 14.4% of sampled ChatGPT responses contained fan-out queries

  • 2026-11-07: Ahrefs corrected its source in October: 14.4% of responses contained fan-out queries, rather than 14.4% searching the web. Our headline and inference that six in seven answers relied only on prior associations were too broad and are corrected. Reddit’s separate figure measures answers citing Reddit, not its share of citation links.

Ahrefs: AI Overviews now appear on 83% of branded searches

  • 2026-10-09: This item originally said 76.2% of "the branded AI Overviews Ahrefs sampled" sat in first position, and that Wikipedia drew 71.9% of citations. Both figures come from a separate Ahrefs panel of 100 bare brand names, not from the 82.91% branded-keyword sample: 63 of those 100 names showed an AI Overview by September 30, 48 of them (76.2%) in first position, and Wikipedia was cited in 41 of the 57 AI Overviews whose sources were visible (71.9%).

ChatGPT-5.6 Luna shifts to site: search, cites fewer sources

  • 2026-09-28: This item originally said Luna used the site: operator in "76% of searches" and that "total search volume held roughly steady." Seer measured fan-out queries — the sub-queries ChatGPT runs while answering — not searches: 76% of Luna's fan-out queries used site:, against 21.6% for GPT-5.5, and the total number of fan-out queries held steady (729 versus 733).

SISTRIX measures weekly churn in AI search’s cited domains

  • 2026-11-02: This item said 86% of prompts have a stable core of a few cited domains. SISTRIX measured that for Google AI Mode only, not for AI Overviews or ChatGPT; the figure is now limited to AI Mode, and the same scope is corrected where the glossary, State, best-practices and numbers pages repeat it.
  • 2026-11-07: Our original summary confused domain churn-in with URL replacement, applied AI Mode’s stable-core findings too broadly, treated Jaccard overlap as a percentage of prompts, and generalized the News/Media domain category to all news articles. The headline, answers and analysis now preserve those distinctions.

Cloudflare syncs robots.txt, sets four conditions for crawlers that also train

  • 2026-08-27: This item originally said Cloudflare "rewrites" a site's robots.txt and that Bot Preference Sync "auto-generates" the file. It does neither. Cloudflare prepends a managed block and, in its own words, any existing Disallow directives are maintained; the sync is opt-in and can be turned off at any time. The headline has been changed from "Cloudflare rewrites robots.txt, gates access on disclosure", which also implied the disclosure conditions apply to crawling generally when they apply only to crawlers that both search and train, seeking access to sites that disallow training. The URL is unchanged.

Study finds 22.4% of tracked text-fragment events assigned to Direct

  • 2026-11-07: We described text-fragment events as AI Overview clicks and sessions. The study counted 51,200 events on one site; fragments can also occur outside AI Overviews. Its 7.53% comparison uses events and organic sessions, so it cannot establish AI Overviews’ share of sessions. The headline, answers and body are corrected.

Ahrefs study: llms.txt files mostly unread by AI crawlers

  • 2026-11-07: The 58% CTR estimate used December 2025 data and was published in February 2026. It came from a separate 300,000-keyword analysis, not the 15-million-point analysis described earlier in this item. The date and attribution are corrected.

Study ties AI citation likelihood to how recently a page was updated

  • 2026-11-07: The 72% versus 42% comparison classifies the same 4,124 cited pages by update date and publication date. It does not show that refreshed old pages outperform newly published ones. The answer and interpretation are corrected.

People Inc. won't block Google's AI crawlers despite traffic slide

  • 2026-11-07: The 21% share refers to Core Sessions, not total traffic. The Q2 investor presentation reports Google Search Core Sessions down 40% year over year and total Core Sessions down 22%. The answer and body now use the company’s metric.
  • 2026-11-07: The reusable social summary still called the 40% decline Google traffic. It now names the measured unit, Google Search Core Sessions, matching the corrected article and Q2 source. Historical social posts are a separate correction surface.

Pillarbase study: 15.7 million AI Mode citations show what Google reuses

  • 2026-11-07: The 117-word median belongs to 528 reconstructed client passages, not all repeatedly cited passages. We also called it an average in the answer. The cohorts and statistic are now distinguished; the study does not establish an optimal passage length.

Search Console adds Generative AI performance reports

  • 2026-11-14: We said Search Console's new generative-AI report gives sites first-party impression and click data for AI surfaces. Google's announcement describes impressions only (with page, country, device and date breakdowns), not clicks, CTR or position, and it began as a limited rollout that reached all sites on August 31, 2026. The answer box above repeats the error.