Market analysis

· 11 min

Marketing analytics in October 2026: measurement now depends on other people's courtesy

One undocumented change in ChatGPT moved a third to a half of its tracked sessions into direct traffic within days. The platforms are adding AI reports that count impressions, not clicks, and the tools that watch Google from outside are being fed altered pages.

Analytics
AI35%GOOGLE & SEARCH18%PPC3%ANALYTICS34%BRANDS10%
Topic radarSelect a topic to filter the archive. The radar shows how this issue's words divide across the topics.

Signal readout

Topic mix

  • AI: 35%
  • Google & Search: 18%
  • PPC: 3%
  • Analytics: 34%
  • Brands: 10%
  • AI35%
  • Google & Search18%
  • PPC3%
  • Analytics34%
  • Brands10%

Share of the issue's words about each topic, counted section by section.

Why it matters

Annotate 7 to 15 October in every report that compares AI referrals across that week, and check direct traffic to the same landing pages before concluding that ChatGPT traffic fell. A tag is a courtesy an assistant can withdraw without notice; a measurement of what the answer says does not depend on it.

The short answer

What
A reading of marketing measurement in mid-October 2026: AI referral tracking and the ChatGPT tagging break, GA4 and Google's measurement stack, Search Console and Bing AI reporting, the AI visibility tools market, bot traffic and the state of rank tracking.
Why
Because the signals analytics relies on are now set by assistants, platforms and crawlers, and several of them failed or changed this quarter. Decisions made on those numbers need to know which ones still hold.
Who
Analysts, marketers who report AI or search performance, and teams buying AI visibility or measurement tools.
Where
Global, with platform features as Google and Microsoft released them and US data where only US data exists.
When
May to 16 October 2026, with the ChatGPT tagging episode of 7–15 October in detail.
How we know
From Google Analytics release notes and Google's measurement blog, Search Console and Bing documentation, vendor data (Adobe, Profound, Graphite, Peec AI, BrightEdge, Chartbeat, Cloudflare, Nozzle), company announcements on funding and acquisitions, and the trade press. Every source is listed at the end.

In plain words

Analytics tools know where a visitor came from mostly because the place that sent them leaves a label. ChatGPT's labels went missing for a week this October, and its visitors looked like people who typed the address by hand. Google and Bing now show how often a site appears in AI answers, but not how many people clicked. And more than half of web traffic is now robots. So the numbers marketers look at depend more and more on choices made by someone else.

Analytics has always depended on other people's labels. A referrer is set by the browser and the site that sent the visit; a UTM tag is added by whoever built the link; a ranking is whatever a crawler was shown. This quarter showed how much of marketing measurement now rests on those courtesies, and how quickly they can be withdrawn.

This analysis starts with the week ChatGPT's traffic went dark, then looks at what Google and Microsoft now report about AI surfaces, the market of tools built to fill the gap, the robots in the logs, and the trackers watching Google from outside. It ends with our reading of what still deserves trust.

The week ChatGPT's traffic went dark

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

Several measurements of the same break agree on size. Adobe's Cedric Huesler, looking at about 500 pages where ChatGPT had supplied at least 5% of visits, saw ChatGPT visits fall 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. Profound measured the share of cited ChatGPT answers carrying a tagged link at about 98% on 8 October, 75% on 9 October and 66% on 14 October, while answers using the new Intelligent UI rose from 2% to about 35%. Graphite saw GA4 sessions attributed to ChatGPT fall 22–57% from 9 October; LLM Pulse saw about 46% fewer ChatGPT sessions with total traffic flat. Search Engine Journal noted that all of these companies sell AI search tools, and that two of them gave no sample size.

On 15 October Huesler reported that OpenAI had told him the missing tag was a known bug, and that by 9am Pacific tagged ChatGPT clicks were back to 95–100% of the same hour a week earlier. OpenAI has made no public statement. The referrer is still missing on most ChatGPT clicks, and the visits recorded as direct between 8 and 15 October will stay there.

MeasureBeforeAfterWho measured
Cited answers with a tagged link≈98% (8 Oct)66% (14 Oct)Profound
ChatGPT visits, ~500 pagesbaseline week−31,000 a dayAdobe (Huesler)
Direct visits, same pagesbaseline week+26,000 a dayAdobe (Huesler)
GA4 sessions attributed to ChatGPTbefore 9 Oct−22% to −57%Graphite

What the platforms now report

GA4 added an AI Assistant channel to its default channel grouping on 13 May. Traffic from recognised chatbot referrers gets the medium ai-assistant; Google's documentation names ChatGPT, Gemini, DeepSeek, Copilot and Grok as examples without publishing the full list, and deliberately leaves AI Overviews and AI Mode in Organic Search. The channel depends on the referrer, which is exactly what ChatGPT clicks are still missing.

The rest of the year's GA4 releases are plumbing: a Data Manager API for server-to-server events (7 May), customisable conversion windows of up to 90 days for clicks (11 August), drag-and-drop dashboards (9 September), hostname filters (21 September). On 10 September Google announced Data Manager inside GA and DV360, enhanced conversions in both, and agentic help in Meridian, its open-source marketing mix model, which can now read brand signals such as branded search volume; GeoX, its library for geo experiments, became generally available inside Meridian. Google's own figures claim higher incremental return for advertisers who connect offline data and a 14% average conversion uplift for its server-side tag gateway.

Search Console's generative AI reports, available to all sites from 31 August, count impressions in AI Overviews, AI Mode and Discover's AI features and show no clicks and no queries; John Mueller said in September that an AI Overview is counted as a block, because a link's position inside it is hard to make useful. Bing's AI Performance report counts citations in Copilot and Bing's AI answers, and since 13 October groups the grounding queries into topics with estimates Microsoft calls "directional signals for planning".

Try it on your own page

See what Voicescope finds on a page of yours

Everything above is measurable on your own work, and the check takes about a minute. Paste one address and read what comes back.

Open Voicescope

Free · no account · one at a time

Position for these is hard to do in a way that makes it useful.

Our reading: every first-party AI report so far counts exposure, not outcome, and none shows what an answer says about a brand. Google's analytics files its own AI surfaces as organic, and its console shows them without clicks. The question "did the answer recommend us" is still answered only outside the platforms.

A market built on the gap

Capital has followed. Profound raised a $180 million Series D at a $1.8 billion valuation on 15 September, less than seven months after a $96 million Series C at $1 billion, and says it serves more than 1,000 enterprise brands. Adobe completed its acquisition of Semrush on 28 April and in June announced Adobe Brand Visibility, combining its LLM Optimizer with Semrush's AI tools. Sitecore bought Scrunch in June. Searchable bought Meridian Tech in September, a merger of two AI visibility start-ups. Peec AI said in May it had passed $10 million in annual recurring revenue 16 months after launch. Similarweb launched AI Ads intelligence on 17 August, estimating that about 26% of ChatGPT Free and Go responses carry a sponsored placement.

Our reading: suite vendors are buying AI answer measurement as a feature, and pure players are consolidating. Buyers should expect methods to differ more than dashboards suggest, and should ask how many prompts, how often, on which engines and with what repetition each number rests on.

More than half the traffic is not people

Cloudflare's founders' letter of 27 September says automated traffic overtook human traffic in May 2026; it had forecast that for the second half of 2027. Its July report put AI training at 52% of crawler requests in June, up from 22% in spring 2025, with mixed-use crawlers at more than 36% and pure search crawling a small and declining share. More than half of AI crawler traffic re-fetches unchanged pages. Cloudflare turned its pay-per-crawl marketplace into pay-per-use, paying publishers when content is used rather than fetched, and from 15 September its defaults block mixed-use crawlers from pages carrying ads for new and free customers.

Publishers are pushing for visibility into that traffic: 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 a coalition including the Guardian, the FT, the BBC and AP published SPUR, a standard for reporting when content is retrieved, used as grounding, cited, shown and clicked.

For analytics this means server logs and sessions now describe agents as much as people. A traffic report that does not separate them measures crawl policy as much as demand.

The trackers watching Google

From about 13 September Google became markedly better at blocking scrapers. Nozzle reported roughly 80% less data, SISTRIX and Semrush slower collection; Ahrefs showed no visible loss. Then Nozzle found altered pages: 8.0% of 17.8 million Google results pages collected from 13 September to 9 October across 49 countries had wrong results on a normal-looking page, against a 0.17% baseline, peaking at 27.0% on 29 September. Google has not commented. In court, a judge dismissed Google's anti-circumvention claims against the scraping service SerpApi on 20 July, holding that plain search results are not protected works; Google filed an amended complaint on 11 August.

Third-party citation data has its own limits. Profound's study of 883,000 pages across seven engines found the median page loses half its citation share 11 days after its peak, and how long a page stays cited on one engine correlates with another at only 0.03 to 0.09. Ahrefs found ChatGPT's most-cited domains reshuffled between March and September. A share-of-voice number without its engine, its date and its repetition count describes a moment, not a position. The 12 October issue has the Nozzle data in detail.

What it adds up to

This section is our reading, not a measurement.

  • Attribution is set by the assistant, not the marketer. One undocumented interface change hid a third to a half of ChatGPT-attributed sessions; the fix exists only as a second-hand report, and the referrer is still gone.
  • Platforms report exposure. Search Console counts AI impressions, Bing counts citations, GA4 counts referrals when they arrive. None measures the content of the answer.
  • Measurement is moving to models and experiments. Meridian with GeoX, branded search as an MMM input, conversion lift tests and server-side tagging are designed for a world where the click path through an AI answer is invisible.
  • Outside observation needs stated method. Since mid-September, a single-source swing in any Google-derived number is a hypothesis until it is replicated with a stated observation count.
  • The traffic that returns is a sliver. Chartbeat measures AI referrals at about 0.01% of publisher traffic while dark social and direct grow. The value of AI visibility will not be found in referral reports.

What to watch

  • Whether OpenAI confirms the tagging bug publicly, and whether the referrer returns to ChatGPT clicks.
  • Whether GA4's AI Assistant channel starts recognising tagged visits that arrive without a referrer, and names Perplexity and Claude.
  • Whether Search Console adds clicks or queries to its AI reports.
  • Whether the altered results Nozzle measured continue, and how trackers adapt.
  • How much publishers actually earn under pay-per-use, and whether crawler operators split mixed-use bots.

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

Questions people ask

Why did ChatGPT traffic drop in GA4 in October 2026?

After GPT-6 reached ChatGPT on 7–8 October, many links lost the utm_source=chatgpt.com tag and had no referrer, so GA4 filed the visits as direct. Graphite measured GA4 sessions attributed to ChatGPT falling 22–57%. On 15 October Adobe's Cedric Huesler reported that OpenAI called it a bug and that tagging was returning.

Does GA4 have an AI channel?

Yes. Since 13 May 2026 GA4's default channel grouping includes an AI Assistant channel for traffic from recognised chatbot referrers, with the medium ai-assistant. Google's own AI Overviews and AI Mode stay in Organic Search, and the full list of recognised assistants is not published.

How much web traffic is bots in 2026?

Cloudflare says automated traffic overtook human traffic in May 2026. Its July data put AI training at 52% of crawler requests in June 2026, up from 22% in spring 2025.

How reliable are AI visibility and share-of-voice tools?

They measure something the platforms do not, but results vary by engine and fade fast. Profound found the median page loses half its citation share 11 days after its peak, with almost no correlation between engines. Ask any tool for its engines, prompt count, repetition and dates.

Sources

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

The product

Watch it on every page, every day

One reading tells you where a page stands today. The product asks the same questions of the same assistants continuously, so a change is something you are told about rather than something you go looking for.

Request an invitation

Invite-only while we keep the readings honest

Author

Olexandrauthor

15+ years in marketing and SEO. Dozens of products, mostly the US market. Research and measurement.

LinkedIn

Pass it on

LinkedInXRedditHacker News

Read in your language

FrançaisDeutschEspañolУкраїнськаPolski

Opens a browser translation of this English article. The original source stays in English.

All issues