I hope you’re reading this after your morning coffee, because the digital marketing industry did not take its foot off the gas when it comes to updates. AI is rewriting search behaviour, organic clicks are tanking, ad platforms are getting… bold, and Google has decided that “agentic actions” are absolutely a thing now.
No fluff. Just the good stuff. Let’s get into it.
BrightEdge’s latest ecommerce analysis confirms what many of us suspected. AI Overviews are becoming a powerful research sidekick, but they are not where people finish the purchase.
AIO visibility surged to 26% in September, then dropped to 9% in October. Only 30% of keywords survived that pullback. Year on year, churn hit 82% and overall coverage was down 56.8%, which tells you how experimental this format still is.
Categories with clear specs and factual comparisons did well. Grocery and Food, TV and Home Theatre and Small Appliances held on to much of their visibility. Categories led by taste and aesthetics, like Furniture, Home Décor and Apparel, saw almost all AI Overview results disappear.
It is possible that with Black Friday coming, Google may have realised that AI Overviews were soaking up too much commercial attention and, by extension, ad revenue at the worst possible time, but nothing official has been announced yet.
What this means for you
Read more
Seer Interactive has been tracking what actually happens to clicks when AI Overviews appear. The short version: it is not pretty.
For informational queries that trigger an AI Overview, organic click-through rates have dropped 61% since mid-2024. Paid CTR on the same queries is down 68%. Even where there is no AI Overview, organic CTR is still down 41% year on year.
So users are not just clicking less when AI occupies the top of the page. They are clicking less across Google in general because more questions are being answered in AI tools, and more discovery is happening on social platforms and marketplaces. That traffic is not quietly waiting to come back. It has already moved.
There is one bright spot. When brands are actually cited in AI Overviews, they do far better. Those that were cited earned 35% more organic clicks and 91% more paid clicks than brands that were not mentioned at all.
Earlier this year, LLM referral traffic looked unstoppable, growing by about 65% from January. Since July, the trend has flipped for some industries. In the B2B industry, LLM traffic is now down 42.6%, and average monthly growth has slowed from 25.1% in 2024 to 10.4% in November 2025.
LLMs still contribute more to organic traffic than they did last year, up from 0.14% to 1.10%, but this is being drowned out by falling organic clicks from Google due to AI Overviews.
There is a structural shift behind this. ChatGPT 5 introduced a router that prefers models which show fewer citations. At the same time, citations have clustered around a small set of sources such as Reddit and Wikipedia. ChatGPT mentions of many sites dropped by around one-third in October, which means usage is up, but outbound traffic is down.
We have also seen small decreases in November across a handful of client accounts. Without any year-on-year LLM data, it is hard to say exactly why. It could be seasonal patterns, fewer clicks overall, fewer citations inside ChatGPT, or even a slight drop in LLM usage.
It is also worth stressing that the headline numbers here come from a study of only 8 B2B sites, so while the drop is concerning and worth watching, it does not yet represent the entire landscape.
Ahrefs analysed more than 43,000 keywords to answer a simple question: once you appear in an AI Overview, do you get to stay there?
In short, not really. Across the dataset, AI Overview content changed 70% of the time from one observation to the next. On average, each AIO lasted just 2.15 days before regenerating. When the content refreshed, 45.5% of citations changed, which means roughly 1 URL was swapped out every time the same query was re-run. Entities behaved in a similar way, with 46% volatility across updates.
The interesting part is that despite all that churn, the meaning of the AI Overview barely moved. Consecutive responses scored 0.95 on semantic similarity. In other words, Google keeps rewriting the surface of the answer, rotating sources and changing examples, without really changing the core takeaway.
There was also no correlation between search volume and stability. High-volume keywords changed just as often as niche ones, which suggests that Google is not caching popular AI Overviews for longer.
Google is quietly moving AI Mode from “answer engine” to “do things for me” assistant.
After launching restaurant reservations earlier this year, Google has now added 3 new agentic actions for users in Search Labs in the United States. AI Mode can now help book event tickets, beauty appointments and wellness appointments. Google AI Pro and Ultra subscribers also get higher usage limits.
In Glenn Gabe’s test, AI Mode spent around 8 to 10 minutes researching concerts in New York, then returned a list of options with prices and direct links to places like Ticketmaster and StubHub. All of this happened with the user asking a single, high-level query.
The direction is clear. Google wants AI Mode to handle complex, multi-step tasks and keep users inside its own ecosystem as they research and book.
Semrush’s latest AI Visibility Index shows a clear split between how ChatGPT and Google AI Mode are evolving.
ChatGPT has been pulling in more brands and more sources over recent months. Mentions rose about 12% in September before dipping in October, which suggests it is experimenting with how often and how widely it names sources. Overall, ChatGPT’s source pool grew by about 80% in October.
Google AI Mode has headed in the opposite direction. Mentions fell 4% over the same period, and its pool of sources grew at a steady 13%, pointing to a smaller, more stable set of trusted sites.
Reddit is the perfect example. ChatGPT’s mentions of Reddit dropped 82%, although it still remained its fourth most common source. In AI Mode, Reddit mentions surged 75% and Reddit became the second most used source behind Amazon.
Across industries, ChatGPT is seeing more brand variety in Consumer Electronics and Technology, while Finance saw a drop. Google AI Mode stayed quite steady across most verticals, with Fashion and Apparel as a notable growth area.
Microsoft has rolled out a new search experience inside Copilot, and it is one of the more publisher-friendly moves we have seen from an AI platform so far.
Copilot now has a dedicated search mode with clearer citations, more clickable links and a full list of references built directly into each response. Citations are visible at the top of the answer, reference lists are far more complete, and Copilot uses navigation cards for more complex queries so users can refine their intent.
Microsoft also says it has improved the underlying search prompt that Copilot uses, which should mean more accurate and relevant results.
Screenshots shared on X show a cleaner interface with a search tab, source cards and a structured references section. Copilot now feels much closer to a hybrid between a search engine and a conversational assistant.
AI search is starting to blur geographic boundaries in ways that classic SEO never did. As LLMs synthesise answers across languages and markets, they often ignore or override the traditional signals that kept regional content separate.
The result is that global English pages frequently become the “default truth”, while local teams see their market-specific content sidelined. This shows up most clearly in search-grounded AI systems like Google’s AI Overviews and Bing’s generative results. The AI pulls from whichever version of your content appears strongest or most complete, then rewrites the output in the user’s language. It looks local on the surface, but the data underneath is often global.
There are three main reasons. AI systems tend to confuse language with location, training data is heavily skewed toward English, and canonical consolidation means even well tagged local pages can be conceptually absorbed into the global entity.
Hreflang still helps with indexing, but it has very little influence on how generative answers are stitched together. In practice, AI follows authority, depth and entity clustering, not your international site structure.
On the paid side, Microsoft Ads has made a simple but important change. Instead of rejecting an entire ad when one asset violates policy, it can now disapprove just the single asset and keep the rest of the ad running. Microsoft has also released a new “conversion lag to 90%” metric that shows how long it takes until 90% of conversions are reported. That gives a much clearer view of attribution delays.
Developers exploring the ChatGPT Android app have spotted components labelled “AdTarget”, “SearchAds” and carousel-style ad units. A few users have also reported seeing early test ads live.
This is a strong signal that ChatGPT, which has become a major discovery tool for many users, is moving toward monetised ad placements. It may introduce search style units, conversational formats or contextual placements that sit alongside AI answers.
If this rolls out at scale, AI chat platforms will no longer be just a top-of-funnel research tool. They will start competing directly with search and social for ad budgets.
Meta has released fresh data on its Native Reels Ads, and the results are strong.
Ads that match the look and feel of organic Reels content deliver 2.3x higher attention and a 5.3x lift in purchase intent. Reels remains Meta’s fastest-growing surface, and users clearly prefer ads that feel integrated into the content stream rather than interrupting it.
Finally, on Google Ads, AI Max is starting to show its teeth.
Advertisers are reporting that AI Max search campaigns blur the line between match types, override keyword intent and expand reach far beyond the expected query set. AI Max leans heavily on broad matching and AI rewriting of queries, which can introduce irrelevant searches into performance and make it harder to see what is really going on.
For some high-volume accounts, this extra reach can be useful. For brands that need tight control, it can be uncomfortable.
AI is reshaping discovery, rewriting user behaviour and changing where conversions happen. Google, Microsoft, Meta and OpenAI are building new ecosystems faster than anyone can keep score. The businesses that stay ahead will be the ones that focus on visibility, clarity and trust across both classic search and AI-powered assistants.
If you want support adapting any of these shifts into your 2025 strategy, you know where to find us.