MPWhat The Shift Is
AI search is a new category, and it is moving faster than anything search has done in twenty years. The platforms change how they choose sources without announcing it. Google ships updates that reach AI Overviews and blue links at the same time. New measurement tools appear monthly, and the research about them contradicts the research from the month before. A business owner cannot follow all of that, and should not have to.
So every Monday morning we read what changed across Google AI Overviews, ChatGPT, Perplexity, Copilot and the studies around them, and we write down the handful of things that matter to a business trying to get found, understood and recommended. Plain English first. The detailed brief underneath for the marketing people. If you have ever wished somebody would just tell you what changed and what to do about it, without a webinar, this is that.
Two labels do the honest work in every issue. Confirmed means a platform documented it or the data is public. Our read means it is our interpretation, which is worth exactly what you paid for it, and we say so.
In Plain English: What Changed and What To Do
The spam update is over, and fake authors are now officially a problem
Google's September spam update, which began on September 24, was marked complete on October 8. It ran for 13 days and 16 hours, right up against the two weeks Google had allowed, while the June and August spam updates each finished in about two days. On October 5, Google also updated its page on helpful content. It now says to avoid deceptive authorship information, that fabricating creator profiles is a form of deception that makes a page untrustworthy, that producing large amounts of AI text without manual oversight shows little to no effort, and that giving credit to sources does not replace the need for original effort.
While an update is rolling out, a traffic dip is weather. Now that this one is done, what you see is closer to a verdict. Google's AI Overviews and AI Mode draw on the same quality signals as the rest of Search, so the same pages that get demoted in the blue links tend to vanish from the AI answers too. And Google has now said in writing that a made-up expert with an AI headshot counts against you, after years of shrugging at fake bylines. If your blog's author has a stock-photo smile and no history anywhere else on the internet, Google has now said out loud what everyone suspected.
Once Search Console has caught up (see below), compare the two weeks after October 8 with the two weeks before September 24. If you lost ground, look at the pages that fell for thin AI text nobody reviewed, invented authors, or pages that mostly summarize other sites. Put a real person on every article: a real name, a real photo, a real bio that links somewhere real. If nothing moved, there is nothing to fix this week, which is allowed.
Your rankings may not have crashed. Your rank tracker may have
On October 10, the rank-tracking company Nozzle published research showing that since September 13, Google has been returning complete, normal-looking result pages to automated trackers with most of the usual sites swapped out. Across 27.5 million result pages in 49 countries, 8 percent were affected, against 0.17 percent before. On September 27 to 29 it was roughly a quarter. On these pages the tracked site dropped out of the top 10 nearly 59 percent of the time, and in most cases it was back within three days. Sistrix, SE Ranking, DataForSEO, Advanced Web Ranking and Morningscore have all posted notices about unreliable data, and Google's own Search Console performance report ran more than 60 hours behind over the weekend before catching up on October 12. Google has not commented on any of it.
If a report told you last month that you fell off page one and then climbed back a day later, there is a good chance nothing happened to you at all. Google appears to be fighting automated scraping, and the tools that measure your rankings collect data the same way the scrapers do, so they got caught in the crossfire. That is our read; Google has said nothing. Either way, a ranking number from a third-party tool is now an estimate with an error bar, and the error bar got wider in September.
For the next few weeks, treat Search Console as the record and your rank tracker as a hint. A drop that disappears within a few days is probably the tool. A drop that persists, and shows up as fewer impressions and clicks in Search Console, is real. Before anyone panics about a number, ask which tool it came from and what day it was collected.
GPT-6 reached every ChatGPT user, and it draws its own charts now
On October 7, OpenAI started rolling GPT-6 out to ChatGPT's paid plans, with free users following the next day. With it comes a feature OpenAI calls Intelligent UI: ChatGPT can answer with visuals, comparison tables, diagrams, calculators and tappable buttons, and the model decides which format fits the question. OpenAI says GPT-6 makes better decisions about when to look something up, and that for questions that need a web search it starts answering 44 percent sooner on average than the previous model. Separately, a test of labels on ChatGPT shopping results, such as best for beginners, was spotted on October 6, though OpenAI has not confirmed it.
ChatGPT now has more than 1.2 billion weekly users, and this is the version of it that most of them will use. When it builds a comparison table or a cost calculator on the spot, it is doing the job of your comparison page or your pricing page, in the chat, from whatever sources it chose. ChatGPT can now draw the table. The question is whether you are in it. The sooner it answers and the better it looks, the fewer people click through to check.
Ask ChatGPT the three questions your customers ask before they buy from you, the way a customer would phrase them. Note whether it builds a comparison, which businesses are in it, and what it gets wrong about yours. Then make sure the specifics it needs, your prices or price ranges, your service area, what makes you different, are stated plainly on your own site in words a machine can quote.
Google built a fast lane for forum and social posts
On October 8, Google published documentation for a UGC Fresh Data Program. Approved platforms can send user-generated content and engagement signals to Google through a dedicated pipeline that is separate from normal crawling and from the Indexing API, so that fresh posts and discussions are processed and updated in Search quickly. Google wants content sent within minutes of being posted, on public pages with stable addresses and structured data, from platforms with high volume and a large user base. Applications are reviewed within six to eight weeks, and Google says the program does not guarantee that content will appear.
This is not for your business website; the eligibility rules describe Reddit-sized platforms. It matters because it shows where Google is pointed: at conversations between people, delivered faster. Forum threads and social posts already feed AI Overviews and AI Mode heavily, and this program makes them fresher. Google wants the conversation about your industry within minutes of it happening, and not from you.
Find the two or three communities where your customers actually talk, whether that is a subreddit, a trade forum or a local group, and read what is being said about businesses like yours. Answer questions there as yourself, with no sales pitch, when you know the answer. When a customer is happy, a post in one of those places now carries more weight than it did a month ago.
Ads are coming to free ChatGPT, and OpenAI says they will not touch the answers
On October 5, OpenAI announced a visual ad format for ChatGPT: image-based ads that show during image generation for people on the Free and Go plans, kept separate from the image being created. Testing starts later this month in the United States with an initial group of advertisers. OpenAI also named a long list of measurement partners for attribution and incrementality, and wrote that advertising does not influence the answers ChatGPT provides.
There are now two lanes in ChatGPT: the ad lane, which you can eventually pay for, and the answer lane, which you cannot. OpenAI is drawing that line in public, which is good news for anyone who earns a recommendation and bad news for anyone hoping to buy one. For most small businesses there is nothing to buy yet; the first round is a hand-picked group of advertisers.
Nothing to purchase this week. If you run paid media, note that ChatGPT ads will have real measurement from day one, and keep the budget conversation for when it opens up. Keep spending the effort on the lane that cannot be bought: accurate, specific, quotable information about what you do and what other people say about it.
Your To-Do List This Week
- Judge rankings by Search Console, not a tracker.Now that the spam update is done, compare the two weeks after October 8 with the two weeks before September 24 using Search Console impressions and clicks, and treat any one-day drop in a rank tracker as suspect.
- Put a real person on every page.Replace invented or anonymous authors with a real name, photo and bio that links somewhere real, and have that person read any AI-assisted text before it stays live.
- Ask ChatGPT what your customers ask.Run the three questions people ask before they buy from you, note whether ChatGPT builds a comparison and whether you are in it, and fix the specifics on your site that it got wrong or could not find.
The Full Brief
The detailed version, for marketing directors and anyone who has to defend a number in a meeting. Every item carries its source. Where we are interpreting, it says so.
September spam update completes; Google rewrites its authorship and effort guidance
The September 2026 spam update began on September 24, 2026 at 09:15 Pacific and Google's status dashboard recorded "The rollout was complete as of October 8, 2026." The listed duration is 13 days and 16 hours. The August 2026 spam update is listed at 2 days and 16 hours and the June 2026 spam update at 2 days and 1 hour. Google Search Status Dashboard
Google's helpful content documentation, last updated October 5, 2026, now tells site owners to "avoid using deceptive authorship information," describes fabricating creator profiles as "a form of deception," and says "Any form of deception makes a page untrustworthy to both users and our automated quality systems." The same page describes "using generative AI to produce large amounts of text without manual oversight or curation" as little to no effort, and states that "Attribution or giving credit to other sources doesn't replace the need for original effort." Google Search Central: Creating helpful, reliable, people-first content
Three spam updates in four months, with the latest running five times longer than the two before it, reads as a wider net rather than a bigger one. The documentation edits, landing mid-rollout and a week after Google's October 1 instruction to fact-check all AI content by hand, describe the same target from the policy side: scaled, unreviewed text with a fabricated byline. Google has not connected the update and the documentation, and we are not claiming it has. The practical effect is the same either way.
With the rollout complete, run the post-update comparison: two weeks after October 8 against two weeks before September 24, by page group, in Search Console once its data is current. For pages that lost visibility, audit for unreviewed AI text, fabricated or anonymous authorship and summary-only content. Give every article a verifiable author with an off-site footprint. Annotate September 24 and October 8 in reporting.
Google is serving decoy result pages to automated trackers
Nozzle published research on October 10, 2026 covering 27.5 million result pages across 49 countries. Since September 13, 8.0 percent of pages were "poisoned," meaning fewer than half of a keyword's usual page-one sites appeared and most returned within days, against 0.17 percent in the August 20 to September 9 baseline. On September 27 to 29 roughly a quarter of pages were affected, and rates have run about 5 to 11 percent a day since October 1. On poisoned pages the tracked site fell out of the top 10 58.6 percent of the time, versus 2.7 percent in the baseline, and 77.9 percent of affected keywords had their usual sites back within three days. Social sites and forums appeared more often on these pages, and some AI Overviews were missing. Nozzle: Poisoned Google SERPs
Search Engine Roundtable reported on October 12 that Sistrix, DataForSEO, SE Ranking, Advanced Web Ranking and Morningscore have posted notices about incomplete or incorrect ranking data, and that Google has made no statement. Separately, the Search Console performance report ran more than 64 hours behind on October 11 and was back to normal on October 12, again with no comment from Google. Search Engine Roundtable: Third-party Google rank tracking struggling
Nozzle attributes the decoy pages to Google's response to AI scraping and grounding, and that is the most plausible explanation we have seen, but it is an inference. What is not an inference: a share of the volatility reported during the spam update, including some of the phases trackers announced, was measurement noise. Rank data from any third-party tool now needs a confidence check, and the tools themselves are saying so.
Reconcile rank-tracker movement against Search Console impressions and clicks before reporting it. Flag single-day drops that recover within three days as suspect. Where a tool offers it, compare a result page against the keyword's usual page-one set over several weeks, not against yesterday. For AI visibility, this is one more reason to measure what the assistants actually say rather than what a scraper saw.
GPT-6 rolls out to all ChatGPT users with Intelligent UI
OpenAI announced on October 7, 2026 that GPT-6 is rolling out to ChatGPT's Plus, Pro, Business and Enterprise plans, with Free and Go the following day. Paid plans use GPT-6 Sol and free plans GPT-6 Luna. Intelligent UI lets ChatGPT answer with visuals, tappable buttons, forms, charts and interactive tools, choosing the format per question, and the model can begin answering while it continues reasoning. OpenAI says GPT-6 makes better decisions about when to look something up and that "GPT-6 Instant starts answering 44% sooner, on average" than GPT-5.6 Instant on questions that need web search. OpenAI: GPT-6 for everyone
ChatGPT's release notes for October 7 list comparisons, diagrams, calculators and games among the interactive elements, and note that Intelligent UI works from Instant through Extra High reasoning but not with the Pro reasoning option. The October 6 entry adds audio file uploads for transcripts and summaries on paid plans. OpenAI: ChatGPT release notes
This is the version of ChatGPT most of its 1.2 billion weekly users will now meet, and its default answer to a buying question is increasingly a table it assembled itself. The sources it chose decide who is in the table. Reported but unconfirmed: a test of descriptive labels on shopping results, such as best for beginners, spotted on October 6. Taken together, the shopping and comparison surface inside ChatGPT keeps thickening, and the attribute data behind it comes from the open web.
Re-run the standard buyer prompts for each client in GPT-6 and record format (table, calculator, prose), inclusion, accuracy and cited sources. Make the attributes a comparison needs (price range, service area, specialties, differentiators) explicit and consistent on the site and in structured data. Update AI visibility baselines, since model changes move citation patterns.
Google documents the UGC Fresh Data Program
Google's documentation, last updated October 8, 2026, describes the program as "a specialized pipeline designed exclusively for trending UGC content" through which approved platforms send content and interaction signals, independent of the Indexing API and organic crawling. Eligibility requires mainly user-generated, social or forum content on public pages with stable URLs, high volume and a significant user base, OAuth 2.0 and validated JSON-LD such as DiscussionForumPosting or SocialMediaPosting, active moderation, and content "ideally sent within minutes" with engagement counters updated within 72 hours. Google responds to applications within 6 to 8 weeks and states "This program doesn't guarantee that content will appear in Search." Google Search Central: UGC Fresh Data Program
Google has been leaning on forum and social content in both classic results and AI answers for two years; this formalizes a direct feed for the largest sources of it and shortens the delay from post to Search. Nozzle's decoy pages, incidentally, showed more social and forum sites. For businesses the implication is indirect but real: community discussion about a category will surface faster, and recent threads will carry more weight in what Google and its AI say.
Map the communities where each client's customers discuss the category. Monitor mentions there weekly, participate transparently where it adds value, and treat earned community mentions as part of the AI visibility plan alongside reviews and press. Platforms that run their own large forums should review the criteria and consider applying.
OpenAI announces visual ads and measurement for ChatGPT
On October 5, 2026, OpenAI introduced image-based ads that show during image generation for users on the Free and Go plans, kept separate from the generated image, with testing beginning later in October in the US with an initial group of advertisers. Measurement covers data integrations (Hightouch, Tealium, LiveRamp), attribution partners including AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin, full-funnel partners Fospha, Measured and INCRMNTAL, geo incrementality tests with Haus, Measured and WorkMagic, and brand suitability pilots with DoubleVerify and Integral Ad Science. OpenAI states that "advertising does not influence the answers ChatGPT provides." OpenAI: New ChatGPT ads format and measurement
The measurement stack is a signal of intent: this is being built as a performance channel with third-party attribution from the start, not an experiment. The explicit separation of ads from answers is the line that matters for AEO. Organic recommendation inside ChatGPT remains earned, and OpenAI has now put that in writing, which gives us something to point to the next time someone asks whether they can pay their way into the answer.
No media action for most clients yet; eligibility is limited to an initial advertiser group. For retail and app clients with paid budgets, log the partner list against the current stack and plan a test when access opens. Keep the organic program unchanged: accurate, specific, well-structured product and service information and third-party mentions.
Questions People Ask
Is the Google September 2026 spam update finished?
Yes. Google's Search Status Dashboard says the rollout was complete as of October 8, 2026. It started on September 24 and ran for 13 days and 16 hours, close to the two-week maximum Google had set, while the June and August 2026 spam updates each finished in about two days.
Does Google penalize fake author names or AI-generated author bios?
Google's helpful content guidance, updated October 5, 2026, says to avoid deceptive authorship information and describes fabricating creator profiles as a form of deception that makes a page untrustworthy to users and to Google's automated systems. Google calls it a signal of a low-quality page. Use a real author with a real name, photo and bio.
Why does my rank tracker show rankings dropping and recovering within a day?
Research from Nozzle published October 10, 2026 found that since mid September Google has been returning normal-looking result pages to automated trackers with most of the usual sites removed. On those pages a tracked site drops out of the top 10 nearly 59 percent of the time and is usually back within three days. Check Search Console: if impressions and clicks held steady, the drop was the tool, not your site.
What changed in ChatGPT with GPT-6?
GPT-6 rolled out to all ChatGPT users on October 7 and 8, 2026. It can answer with charts, comparison tables, calculators and interactive elements, choosing the format for each question, and OpenAI says it starts answering 44 percent sooner on questions that need a web search. For a business, that means ChatGPT is more likely to build a comparison on the spot, from sources it picks.
Can I pay to be recommended by ChatGPT?
No. OpenAI announced visual ads for ChatGPT's Free and Go plans on October 5, 2026, shown during image generation and tested first with a small group of US advertisers, and wrote that advertising does not influence the answers ChatGPT provides. Being recommended in an answer is earned through accurate, specific information on your site and mentions from other people.
What is The Shift?
The Shift is Agency AEO's weekly brief on AI search news. Every Monday morning we review what changed across Google AI Overviews, ChatGPT, Perplexity, Copilot and the research around them, explain in plain English what it means for a business, and separate what is confirmed from what is our read. It exists because AI search is a new category that changes weekly.
Sources & Further Reading
- Google Search Status Dashboard: September 2026 spam update (complete October 8, 2026)
- Google Search Central: Creating helpful, reliable, people-first content (updated October 5, 2026)
- Google Search Central: UGC Fresh Data Program (October 8, 2026)
- Google Search Central: Documentation updates
- Nozzle: Poisoned Google SERPs (October 10, 2026)
- Search Engine Roundtable: Third-party Google rank tracking struggling now (October 12, 2026)
- Search Engine Roundtable: Google Search Console performance report extremely delayed (October 11, 2026)
- OpenAI: GPT-6 for everyone (October 7, 2026)
- OpenAI: ChatGPT release notes
- OpenAI: New ChatGPT ads format and measurement (October 5, 2026)
- Search Engine Roundtable: OpenAI tests new product labels on ChatGPT results (October 6, 2026)