Marketing analytics in October 2026: the month the labels came off

Analytics has always run on other people's labels. A referrer is set by the browser and the site that sent the visit, a UTM tag by whoever built the link, a ranking by whatever a crawler was shown, a backlink count by whoever keeps the index. In October 2026 almost every one of those labels came off at least once.

The mid-month edition of this analysis, published on 16 October, described measurement that "depends on other people's courtesy". The second half of the month made the case harder and more useful. Gemini's tags turned out to have stopped weeks earlier. Google said its fastest-growing search surface earns less per search, and its ad reports still cannot separate it. Two careful studies put numbers on how unreliable a single AI-visibility check is. And the third-party data that SEO teams read every day, rankings and backlinks, shifted for reasons that had nothing to do with the sites being measured.

This analysis takes the month by theme: the referral labels, how small the measurable AI stream is, what the platforms report, how counting is moving to the platforms' side, how to measure AI visibility properly, the robots in the logs, and the third-party indexes. It ends with our reading and what to watch in November.

Two assistants, two broken labels

ChatGPT has tagged links in its answers with utm_source=chatgpt.com since June 2025, which is what made AI referral traffic countable at all. After GPT-6 reached ChatGPT on 7 October for paid plans and 8 October for Free and Go, many links arrived with neither the tag nor a referrer, and analytics filed the visits as direct.

Huesler's data showed where the visitors went. On about 500 pages where ChatGPT had driven at least 5% of visits, ChatGPT visits fell by about 31,000 a day between 9 and 13 October against the week before, while direct visits to the same pages rose by about 26,000. Direct traffic rose on two of every three pages that lost ChatGPT traffic. Graphite saw GA4 sessions attributed to ChatGPT fall 22–57% from 9 October. Profound, which collects ChatGPT answers, tied the change to Intelligent UI, GPT-6's visual answer format, and published the share of cited answers that still carried a tagged link. Profound gave no sample size.

It was a bug. Huesler said OpenAI confirmed it to him on 14 October and began rolling out a fix early on 15 October; by 9 a.m. Pacific that day tagged ChatGPT clicks were back to about 95–100% of the same hour a week earlier, across thousands of domains. OpenAI's release notes do not mention it, so the confirmation is second-hand, and as Profound noted, attribution already lost will not come back. Every report that compares AI referrals across 8–15 October undercounts ChatGPT and overcounts direct.

Gemini's story runs the other way and is murkier. On 1 October Search Engine Journal reported that Gemini had started adding UTM parameters to its outbound links, which Google never documented; John Mueller said only that he saw them too. The Reddit thread that found them said the Gemini mobile app passes a referrer only about 9% of the time. Then on 30 October Search Engine Roundtable reported that Gemini had stopped adding utm_source=gemini around 28 September, after starting in early September. The desktop web still passes a referrer; the iOS app sends none, so those visits land in direct. Google has confirmed none of it. Put the two reports side by side and the tags were already disappearing as the first one was being written.

One more false alarm belongs in the record: on 2 October a report of tracking parameters on AI Overview links was withdrawn. It was a Chrome extension.

The measurable AI stream is small, and getting harder to see

While the labels were breaking, three studies published in the same week of October asked how much traffic AI assistants send at all. The most detailed came from Marie Haynes on 21 October, under the title "GEO is a distraction": 12 months of GA4 data across 100 websites and more than 145 million sessions.

The visitors ChatGPT sent converted better, at 13.8% against 3.02% for Google organic, but Google still drove 96.5% of all conversions across the channels she analysed. SEO Works found organic search delivered 39 times the traffic of AI assistants and 3.2 AI conversions for every 100 organic ones. Kevin Indig, working with 12 sites, wrote that he could not "yet find a reliable relationship between AI Visibility and branded search or direct traffic." Chartbeat, as quoted in mid-month, measures AI referrals at about 0.01% of publisher traffic. The samples are small, and an absence of measurable lift is not proof that none exists.

Search itself is a shrinking share of what publishers see. Chartbeat's Q3 2026 report, published on 15 October, puts search at 12% of North American publisher page views against 21% a year earlier, with no absolute numbers and no cause named. Reddit told shareholders on 29 October that "search traffic has dropped by more than 1 million referrals on average per week" during October. And seoClarity estimated on 28 October that ChatGPT's search-like activity is "approximately 7.5% of Google's search-query volume" — a model built on OpenAI's own message data, not a count.

Our reading: put the label breakage next to these numbers and the referral report is both small and leaky. It is a fair measure of what AI assistants send; it was never a measure of what they say. A brand can be named in an answer without a click, and in October a click could arrive without a name.

What the platforms report: exposure, not outcome

GA4 has had an AI Assistant channel in its default channel grouping since 13 May. It recognises chatbot referrers — Google names ChatGPT, Gemini, DeepSeek, Copilot and Grok as examples without publishing the list — and leaves AI Overviews and AI Mode in Organic Search. It depends on the referrer, which is exactly what ChatGPT's broken links and Gemini's iOS app do not send.

Search Console's generative AI reports, open to all sites since 31 August, count impressions in AI Overviews, AI Mode and Discover's AI features, with no clicks and no queries. October did not change that, and two Google announcements showed why it matters. On 15 October Google said it had made "continued and substantial increases" to inline links in AI Mode and AI Overviews and would put them higher more often, but published no click data; in Search Console every link in an AI Overview takes the Overview's position, so a link that moves up inside the answer reports the same position as before. And on 28 October Alphabet's chief business officer Philipp Schindler said the strong growth in AI Mode queries had weighed on Search revenue "because of the lower monetization". Google Ads reports AI Mode inside ordinary Search and Shopping campaigns, so an advertiser cannot see what an AI Mode click costs or converts at.

Search Console itself wobbled. On 11 October its Performance report ran more than 64 hours behind, a delay not seen since December 2025; it was current again the next morning. Later in the month the Page Indexing report stopped updating after 16 October, and John Mueller said that was nothing to worry about: "Indexing doesn't change quickly, and when it does, you see that reflected in the performance reports first." The useful changes were small: filters for several countries and devices at once (6 October), and a notice on the Search generative AI control page when a child property overrides the parent's setting.

Microsoft went further than Google on AI. On 13 October it added Topics Portfolio, in beta, to Bing Webmaster Tools' AI Performance report: it groups the queries a site is cited for in Copilot and Bing's AI answers into topics, estimates each topic's size, and lists queries where other sites are cited and yours is not. It is the first search-engine report that puts a site's AI citations beside its competitors'. Microsoft's own caveat is the right one.

On the ads side, Google Ads API v26, released on 14 October, adds read-only Search Lift reporting with 15 new metrics, according to PPC Land, and Microsoft said on 28 October that search term landing page reporting will roll out from mid-November. Both measure what an ad caused rather than what it touched. Neither reaches inside an AI answer.

Our reading: every first-party AI report of October counts exposure — impressions, citations, positions of a block — and none reports what the answer said about a brand or whether anyone acted on it. The question "did the answer recommend us" is still answered only outside the platforms.

Counting moves to the platforms' side

The month's quietest measurement change will move the most numbers. PPC Land reported on 28 October that Meta switches on device-side conversion counting by default on 12 November, under a setting called "Measure conversions on customer devices"; advertisers who do not want it have to opt out. It lands about two weeks before Black Friday, so a jump in Meta-reported conversions will arrive exactly when budgets are being raised.

Browsers are closing other doors. WebKit said on 19 October that Safari 27 blocks network access to third parties that turn email addresses into persistent tracking IDs, without naming companies. And platforms are taking more of the ad itself: OpenAI's creative enhancements, on by default for self-serve ChatGPT campaigns from 29 or 30 October, can change the destination URL to another page of the advertiser's site, so landing-page reports and conversion tags have to cover every page the system might choose.

The direction matches what Google built earlier in the year — Meridian, its open-source marketing mix model, with GeoX for geo experiments, server-side tagging and offline data uploads, described in the mid-month edition. The click path through an AI answer is becoming invisible, and the platforms' answer is to count on their own side and model the rest. That is defensible, but it means the advertiser's check on the platform's number has to come from the advertiser's own data.

How to measure AI visibility properly

The AI visibility market kept growing on the gap the platforms leave — Profound's $180 million Series D in September, Adobe's Semrush deal, Sitecore's purchase of Scrunch, Searchable's of Meridian, all covered at mid-month. October's contribution was more valuable than another funding round: evidence on how such numbers should be built.

It started with decay. Profound's study of 883,000 pages across seven engines, published on 30 September, found that the median page loses half its citation share within 11 days of its peak, and that how long a page stays cited on one engine correlates with another at only 0.03 to 0.09. A preprint posted on 11 October, covered by Search Engine Journal on 18 October, ran 80 questions 20 times each on two ChatGPT models: 40.8% and 44.5% of the pages retrieved appeared in only one of the 20 runs, and Google's organic results contained only 6.4–9.4% of the pages ChatGPT cited.

Then, at the end of the month, two studies a day apart answered the practical question. Ahrefs, on 29 October, asked five AI tools the same questions every day for 20 days and followed every site cited on the first day — about 37,500 citations. Half or more were gone the next day.

Ahrefs' advice is blunt: "Don't judge AI visibility from one check." Instead, "Count how many of the last seven days you were cited." It also found that on AI Overviews and AI Mode, pages that ranked anywhere in Google for the same question "were about seven times more likely" to keep their citations for at least a week.

SISTRIX, on 28 October, asked how many questions a measurement needs. It simulated a brand whose true share of answers is 25%. One prompt asked on 20 days could land "anywhere between 0 and 59 per cent"; 20 prompts on one day between 16 and 35; 100 prompts between 21 and 29.

Two more SISTRIX findings belong in every client report. Of all brands that appeared in a prompt's answers, "70 per cent show up on fewer than one in five days." And prompt sets picked by the brand flatter it: "the brand performs better in its own set in three out of four cases, usually by a clear margin."

A third result says the numbers can be moved from outside. A preprint posted on 8 October and covered on 19 October planted invented subjects on sites AI search tools already cite, such as Medium and YouTube; within seven days eight of ten tools were citing the posts. A brand's citation share can shift because of what strangers post, so a sudden citation from a site nobody in the market reads is worth checking. Graphite added on 16 October that brands ChatGPT rarely names from memory reach only 11% median visibility with search even when they appear on the cited pages — being on the page is not the same as being named.

Put together, these studies give a buyer's checklist.

More than half the traffic is not people

Cloudflare's founders' letter of 27 September said automated traffic overtook human traffic in May 2026, and Cloudflare reports that daily requests from AI agents on its network grew by more than 1,700% over the past year. Any log or session report that does not separate agents from people now measures crawl policy as much as demand.

The agents are also harder to tell apart. Gary Illyes said on a podcast, reported on 30 October, that AI made crawler impersonation worse and that firewalls and CDNs "are not perfect" at spotting it. A visit that says it is Googlebot has to be verified by reverse DNS, not trusted by its user agent. And not every agent behaves like a crawler: on 9 October Anthropic published cases from its own evaluations of models submitting real forms on live websites and using URL shorteners to get around fetch limits — the kind of visit that can end up in a conversion report.

Who may fetch what kept shifting. Cloudflare and OpenAI said on 15 October that signals from about 1% of Cloudflare's network helped OpenAI "discover and index hundreds of millions of new web pages" for ChatGPT search. Meta's revised crawl terms, which let it crawl sites associated with any business that uses its products, were postponed from 30 October to 20 November with the crawl clause intact. On 21 October Google's Gary Illyes called robots.txt "an access preference protocol": it decides whether a page is fetched, not what is done with it afterwards. Publishers asked for a way to count: more than 300 of them lobbied Congress at the end of September for a law requiring AI crawlers to identify themselves, and on 2 October the SPUR coalition published a standard for reporting when content is retrieved, used as grounding, cited, shown and clicked.

The third-party indexes moved under their users

October was the month when data bought from outside changed for reasons that had nothing to do with the site being measured.

Rank trackers came first. Nozzle's report of 10 October found 8.0% of 17.8 million Google results pages, collected from 13 September to 9 October across 49 countries, altered, against a 0.17% baseline, with a peak of 27.0% on 29 September and a daily share of 4.2–11.4% since 1 October. An altered page looks normal: in a normal result 88.5% of the sites from the previous check are still there, in an altered one 20.7%, and AI Overviews appeared on 34.9% of altered pages against 54.2% of normal ones — so a tracker taking them at face value would have undercounted AI Overviews too. Nozzle's update of 21 October, with data to that day, tells the same story at a larger scale: counting page 1 alone, 8.7% of results pages since 13 September were altered against 0.17% before; counting page 1 or 2, 22.6%, with a peak of 43.5% on 30 September and 35.7% on 21 October, and AI Overview sources on 48.2% of altered pages against 53.6% of normal ones. Google has not commented. SISTRIX's incident for reduced Google data collection, open since 27 August, still said in late September that Google returns "incomplete or distorted" results in some cases. Semrush wrote about "poisoned" results pages on 27 October and said "Google hasn't commented publicly."

The month was full of ranking movement nobody confirmed. Search Engine Roundtable reported tracker volatility on 16 October, on 20 and 21 October, on 24 October and from 27 October, and Google's Search Status Dashboard lists nothing after the September spam update that ended on 8 October. Glenn Gabe noted that Ahrefs and Semrush placed the 20–21 October drops on the wrong dates. In that state a tracker line is a hypothesis to check against Search Console, not a finding.

Then the backlink index moved. On 26 October Majestic said it had removed a wave of spam built to get into backlink tools themselves: more than 10,000 sites deployed to insert adverts for link building services, and "the number of backlink URLs removed from the index is in the hundreds of billions." Many sites will see their referring domain counts drop and Trust Flow shift, though "the core Trust Flow mechanics remain unchanged." Asked three days later about spammy links that a third-party tool shows and Search Console does not, John Mueller said: "I wouldn't anchor on whether Search Console shows them."

What it adds up to

This section is our reading, not a measurement.

• Labels are a courtesy, and both big assistants withdrew it in the same month. ChatGPT for a week, by mistake; Gemini for weeks, without a word. An AI segment built on one signal will break again.
• The measurable stream is small. ChatGPT sent 0.62% of Google organic's sessions in Haynes' 100 sites. The value of being named in an answer will not be found in referral reports, and neither will its absence.
• Platforms report exposure; outcome is modelled. Search Console counts AI impressions, Bing counts citations, Google Ads folds AI Mode into Search, Meta counts conversions on the device. The advertiser's check has to be its own data.
• AI visibility is a statistic, not a snapshot. October's best evidence says one prompt carries ±34 points of error, citations churn daily, and engines barely agree. Prompt count, days and engine belong beside every number.
• Outside data needs an outside check. Poisoned SERPs, unconfirmed volatility and Majestic's purge all changed third-party numbers without anything changing on the sites measured. The more of the picture you collect yourself, the less of it can move under you.

Our tool: measure from your own side

Each of October's measurement problems points the same way. Referrals went dark because someone else's label disappeared; rankings and backlink counts moved because someone else's index changed, as Majestic's purge showed; and Ahrefs and SISTRIX found that one look at AI answers proves little. The answer is data you collect yourself: your own crawl of your own site, your own Search Console export, AI-readiness checked page by page, and AI visibility judged over many questions rather than one check.

That is what GLLmolfar is for. It is GetLoopLoop's desktop crawler, in preview, for Windows and Mac, and it crawls your site on your own computer. It sorts findings by impact — indexability, redirects, canonicals, internal links, structured data — and checks how pages read to AI search: an AI-readiness score, which AI bots may fetch what under robots.txt, and llms.txt. It groups pages by meaning with an on-device embedding model, for topics, cannibalization and semantic search, and does not send the crawl to a cloud AI service. It reads Search Console data read-only, from Google straight to your computer, runs planned crawls with alerts, compares crawls over time, and exports to Excel, Google Sheets and Jira. It sends getlooploop.com only anonymous usage statistics, which can be switched off. And it runs an MCP server, so an AI agent such as Claude Code or Codex can drive it.

What to watch in November

• 12 November: Meta's on-device conversion counting goes on by default. Watch Meta-reported conversions against your own orders.
• Gemini's tags. Whether Google confirms what happened to utm_source=gemini, and whether the iOS app ever sends a referrer.
• ChatGPT's referrer. The tag came back on 15 October; whether OpenAI says anything publicly, and whether the next interface change breaks it again.
• Search Console. Whether its AI reports gain clicks or queries, and whether the Page Indexing report resumes normal updates.
• Mid-November, Microsoft: search term landing page reporting begins rolling out in Microsoft Advertising.
• 20 November, Meta's crawl terms take effect unless postponed again. Check which Meta agents your robots.txt names.
• Rank trackers and link indexes. Whether altered results continue, whether Google confirms any of October's volatility, and how far referring domain counts settle after Majestic's purge.
• AI-visibility vendors. Whether tools start publishing prompt counts and confidence ranges, as SISTRIX's numbers imply they should.

To see what assistants actually say about a brand rather than what reaches analytics, Brandscope reads the answers directly.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named.

• Marketing analytics in October 2026: a market analysis — GetLoopLoop, 16 October 2026
• ChatGPT analytics update: UTM parameters — Search Engine Roundtable, 16 June 2025
• ChatGPT Gets GPT-6 And Intelligent UI For Interactive Answers — Search Engine Journal, 7 October 2026
• OpenAI Drops UTM Parameters From ChatGPT Results (GPT-6) — Search Engine Roundtable, on Cedric Huesler's data, 14 October 2026, updated 15 October
• ChatGPT's missing UTM tags: what our research shows — Profound, 14 October 2026
• ChatGPT Traffic Drop Appears To Be A Tracking Problem — Search Engine Journal, 15 October 2026
• Google Gemini Adds UTM Parameters For Referral Attribution — Search Engine Journal, 1 October 2026
• Gemini UTM parameters gone — Search Engine Roundtable, 30 October 2026
• Google AI Overview link tracking parameters (corrected) — Search Engine Roundtable, 2 October 2026
• 3 New Studies Suggest AI Visibility & Search Not Worth Investing In — Search Engine Roundtable, 22 October 2026
• GEO is a distraction — Marie Haynes, 21 October 2026
• Google Search's decline for publishers — Search Engine Journal on Chartbeat, 24 September 2026
• Search Share Fell To 12% For North American Publishers: Chartbeat — Search Engine Journal, 20 October 2026
• Reddit ad revenue gains 54% to $847m as daily users stall — PPC Land, 29 October 2026
• ChatGPT vs Google search volume — seoClarity, 28 October 2026
• Default channel group definitions — Google Analytics Help
• Search Console AI reports rolled out worldwide — Search Engine Journal, 31 August 2026
• More ways to connect with the web from Search — Google, 15 October 2026
• Google Puts Inline Links Higher In More AI Search Responses — Search Engine Journal, 15 October 2026
• Google: AI Mode ads earn less — Search Engine Roundtable, 29 October 2026
• Google's ad growth slows to 10.5% as AI Mode ads go to 10 more markets — PPC Land, 29 October 2026
• Google Search Console Performance Report Extremely Delayed — Search Engine Roundtable, 11 October 2026
• Google responds to concern over Search Console indexing reports — Search Engine Journal, 27 October 2026
• Google Search Console Lets You Filter By Multiple Countries — Search Engine Journal, 6 October 2026
• Google Search Generative AI Control Page Adds Top-Level Domain Details — Search Engine Roundtable, 6 October 2026
• Topics Portfolio (beta) in AI Performance — Bing Webmaster Blog, 13 October 2026
• Google Ads API v26 adds 15 Search Lift metrics to measure ad-driven searches — PPC Land, 18 October 2026
• What's new in Microsoft Advertising this October — Microsoft Advertising, 28 October 2026
• Meta Pixel counts more sales from Nov 12 as AI ads go undisclosed — PPC Land, 28 October 2026
• Safari 27 blocks identity firms that turn email addresses into tracking IDs — PPC Land, on WebKit's post of 19 October 2026
• OpenAI to rewrite ChatGPT ads by default as advertisers get days to opt out — PPC Land, 21 October 2026
• Data strength updates — Google Ads & Commerce Blog, 10 September 2026
• Profound raises $180M Series D — Profound, 15 September 2026
• The half-life of an AI citation is 11 days — Profound, 30 September 2026
• Understanding ChatGPT Citations: Evidence from Production Web Search Across Model Generations — arXiv preprint, 11 October 2026
• OpenAI's API Only Partly Matched ChatGPT's Citations, Paper Finds — Search Engine Journal, 18 October 2026
• How stable are AI citations? — Ahrefs, 29 October 2026
• Measuring AI visibility — SISTRIX, 28 October 2026
• From Public Posts to AI-Search Citations: Measuring the Fragility of AI Search — arXiv preprint, 8 October 2026
• AI Search Tools Cited A Made-Up Concept Within A Week, Paper Finds — Search Engine Journal, 19 October 2026
• Parametric knowledge optimization — Graphite, 16 October 2026
• Cloudflare's 2026 annual founders' letter — Cloudflare, 27 September 2026
• Google Is Fighting SERP Tracking. AI Agents May Make It Worse — Search Engine Journal, 5 October 2026
• Google: AI made bot impersonation worse — Search Engine Roundtable, 30 October 2026
• Investigating unintended model actions — Anthropic, 9 October 2026
• Cloudflare and OpenAI: making AI discovery better for everyone — Cloudflare, 15 October 2026
• Meta Delays The Terms That Let It Crawl Business Sites — Search Engine Journal, 26 October 2026
• Google says robots.txt limits AI crawling, not what bots do with data — PPC Land, 21 October 2026
• Publishers ask Congress to force AI bots to say who they are — PPC Land, 29 September 2026
• SPUR publishes AI content tracking standard — Digiday, 2 October 2026
• Poisoned Google SERPs — Nozzle, 10 October 2026, updated 21 October
• SISTRIX status — SISTRIX, read 6 October 2026
• Poisoned SERPs — Semrush, 27 October 2026
• Third-Party Google Rank Tracking Struggling Now — Search Engine Roundtable, 12 October 2026
• Google Search Ranking Volatility & Chatter Spike Today — Search Engine Roundtable, 16 October 2026
• Google Search Ranking Volatility & Chatter Heats Up October 20 & 21 — Search Engine Roundtable, 21 October 2026
• Google Search Ranking Volatility Gets Heated October 24th — Search Engine Roundtable, 25 October 2026
• Google search ranking volatility — Search Engine Roundtable, 27 and 29 October 2026
• Google Search Status Dashboard — Google
• Fighting referring domain spam — Majestic, 26 October 2026
• Search Console does not show all links — Search Engine Roundtable, 29 October 2026
• GLLmolfar — GetLoopLoop, desktop crawler, preview