February 2026: Discover Update, Listicles Hit & the Agentic Search Push Accelerates

February 2026 felt like one of those months where several separate Google stories all pointed in the same direction. On the surface, we had an unconfirmed hit to self-promotional “best X” pages, a confirmed Discover core update, another very strong Alphabet earnings report, and a string of announcements around WebMCP, UCP and AI-mode commerce. Underneath it all, though, the signal was consistent: Google is getting stricter about low-value content and more serious about keeping discovery, decision-making and checkout inside its own AI-led interfaces.

Day of reckoning as Google targets self-promotional GEO listicles

One of the biggest talking points this month was the apparent hit to self-promotional listicles, especially in SaaS. Many of the “best X” pages which had surged on the back of GEO-style tactics appear to have been hit hard in recent unconfirmed updates. Google has not confirmed a dedicated update for this, but the pattern is strong enough that most SEOs will recognise what happened when they look at the winners and losers.

The important point is that this does not look like Google declaring war on listicles as a format. It looks more like Google getting better at isolating a specific subtype: pages that pose as neutral comparisons while being obviously self-serving, thin, repetitive, and engineered to rank across large numbers of near-duplicate commercial queries. That distinction matters, because it means the tactic is not dead; the lazy version of it is.

My Take: I’ve been saying to clients for many months that this day would come. Only the strongest, most valuable, most transparent and most generous listicles have survived — exactly as we predicted. As usual, a lot of the press around this has been sensationalist and lacking nuance, but the impact was undeniably massive for those sites that were most heavily affected.

It’s interesting that Google didn’t announce this, or even confirm it. Perhaps they didn’t want to admit there was ever a GEO loophole so obvious in the first place. Nothing about the way this is working is actually new. It seems to me this is just a recalibration of some older systems to make them isolate and target a specific kind of spam. The ‘Reviews’ system, previously the Product Reviews system, likely played a role in highlighting low-quality product or brand summaries lacking first-hand experience, based around consensus signals and pattern-matching. There is also a strong whiff of the often forgotten unconfirmed Maccabees updates from late 2017 too. Those targeted affiliate listicles, doorway pages and low-quality, near-duplicate and thin-content site areas designed to target keyword permutations. Combine those two with a drop of SpamBrain pattern-matching, trained on the obvious footprints of the worst offenders with hundreds of GEO-targeted near-duplicate listicles, and this recent move by Google suddenly seems not just inevitable but actually a doddle to implement — you almost have to ask why it took so long.

To be clear, and to say the part most of the press around this update either fails to recognise or is too nervous to suggest, listicles can still work for GEO. They work when the listicle is genuinely helpful, is unbiased, and is not plainly self-promotional while critical of every competitor. Here are the rules to stick to:

• Use clear evaluation criteria.
• Be transparently self-critical.
• Be generous to the competitors in your list.
• Don’t stray too far from the consensus around what the ‘best X’ looks like.
• If two listicles will contain the same top 10, or top X, then roll them up into one listicle by combining the criteria. You can only justify another listicle if the competitors you include are justifiably different.
• Most importantly, write a helpful, unbiased user guide which genuinely helps people decide between the best options for what they need. Remember, every single inclusion in your top 10, or top X, is there because they do solve a need very well for a certain type of customer, so don’t be afraid to add persona-based context. This brand is best for people who need an easy entry point. This brand is best for expert users. This brand is best if you want X, Y and Z.

Google’s February 2026 Discover core update puts editorial quality under the microscope

Google officially launched the February 2026 Discover core update on 5 February 2026 and confirmed the rollout completed on 27 February 2026. Google described it as a broad update to its systems that surface articles in Discover. At launch, it applied to English-language users in the US, with plans to expand more broadly over time.

The stated aim was to reduce clickbait and sensational content, surface more in-depth, original and timely material, and improve localisation. John Mueller’s accompanying comment was revealing. He said Google’s systems are designed to identify expertise on a topic-by-topic basis, meaning broad publishers are not excluded so long as they demonstrate real strength in the specific subject area they are covering.

There was also a wider backdrop of unusual volatility. While Google framed this as a Discover-focused update, ranking volatility across search remained elevated through late February and into early March. That suggests there may have been a broader ripple effect, whether by design or coincidence.

My Take: This boils down to two words: topical authority.

The Discover angle sharpens that idea rather than changing it. It is not only about proving you know a subject. It is also about proving you can cover it in a way that is balanced, insightful, timely and non-sensational. Google is effectively rewarding publishers that behave more like editors and less like growth hackers.

That matters beyond Discover. For clients building visibility in AI search, news-led content still has huge strategic value — for Discover, for branded search growth, for citations in AI systems, and for overall authority-building. But the bar is rising. If you are going to publish news-style content, it needs to feel like something written by people who understand the beat, not like a thin reaction piece rushed out to capture a trend. The winners here are the publishers who can add judgement, not just speed.

Alphabet’s Q4 2025 results show Google Search is still a commercial giant

Alphabet’s fourth-quarter 2025 results were, by any normal standard, enormous. The company said annual revenue topped $400 billion for the first time. Search and advertising remained exceptionally strong, with Search showing continued momentum. Sundar Pichai said: “Search saw more usage than ever before, with AI continuing to drive an expansionary moment.” Alphabet also said its first-party models now process more than 10 billion tokens per minute via direct API use, and that the Gemini app had grown to over 750 million monthly active users.

This is a useful reality check for anyone too eager to write Google’s obituary. Yes, AI-native products have changed user expectations. Yes, ChatGPT and others have forced Google to move faster. But Google is still operating from an absurdly powerful starting point: massive distribution, default user behaviour, and the highest-intent commercial inventory on the web.

My Take: The idea that Google is under threat from the likes of ChatGPT isn’t entirely unfounded. However, I think it would be more accurate to say the likes of ChatGPT are merely keeping Google on their toes.

Their business is strong and their ownership of the search market alone is more than enough to make their AI models the most visible for the average person. Even without Gemini, and without doing anything more than bolting AI answers onto its existing search platform, Google is doing enough to keep pace. From a revenue perspective, the commercial intent is much higher in search than in ChatGPT and therefore the ad revenue potential is much greater for Google.

As we’ve predicted many times before, while we love geeking out over model capabilities as much as the next tech nerd, it’s likely the winner of the AI race won’t necessarily be determined by the model, but by its reach — and Google’s monopoly makes them a pretty safe bet.

WebMCP and UCP-powered checkout make Google’s agentic ambitions impossible to ignore

Two stories this month made Google’s direction of travel very obvious.

First, Google’s Chrome team announced an early preview of WebMCP, a standard way of exposing structured tools so AI agents can perform actions on websites with greater speed, reliability and precision. In plain English, WebMCP is about making websites easier for agents to understand and act on without brittle guesswork.

Second, Google pushed further on commerce by launching UCP-powered checkout in AI Mode and publishing supporting Merchant Center documentation. UCP, the Universal Commerce Protocol, is being positioned as an open standard for agentic commerce that can work across the whole shopping journey.

That combination matters. WebMCP is about making actions machine-readable. UCP is about turning those actions into transactions. Together, they point towards a search journey where discovery, comparison and checkout are increasingly mediated by agents rather than by ten blue links and a shopping basket flow the user manually controls.

My Take: Clear intent from Google: the future of search is agentic.

For SEO and ecommerce teams, that means the optimisation surface is expanding. It is no longer just about ranking a page and persuading a human. It is increasingly about making your content, product data and site actions legible to systems that may decide, compare and even transact on a user’s behalf. The commercial opportunity here is obvious, but so is the risk: if Google owns more of the journey, brands may gain efficiency while losing some direct customer interaction and some control over the path to conversion.

Google’s AI-generated landing page patent offers a glimpse of a more disintermediated web

Another eye-catching development was a newly surfaced Google patent for an AI-generated content page tailored to a specific user. The patent describes a system in which Google could generate a custom page aligned to a user’s intent and send searchers there instead of to the site’s standard landing page.

As always with patents, caution is required. A patent is not a product launch, and many patented ideas never become mainstream features. Still, patents are useful because they show where a company’s thinking is heading. In this case, the direction is unmistakable: more dynamic mediation between query and destination, and potentially less dependence on the exact page a publisher has chosen to present.

My Take: If AI Overviews made publishers nervous, this patent will not calm anyone down.

Even if it never ships in full, it captures a very real strategic possibility: Google deciding that the best landing page for a user is one it assembles itself. For brands, that would raise serious questions about attribution, UX control, messaging consistency and what it even means to own the visit. It is another reminder that the long-term challenge in search is no longer just ranking. It is staying meaningfully present in a journey that platforms increasingly want to orchestrate themselves.

OpenAI begins testing ads in ChatGPT

Away from Google, OpenAI formally began testing ads in ChatGPT in the US in February. The initial rollout is for logged-in adult users on the Free and Go tiers, with paid and enterprise tiers excluded.

This matters not because the first ad format is revolutionary, but because it confirms the broader direction of travel. AI assistants are not only answer engines; they are becoming monetised discovery environments.

My Take: This was always coming.

Once AI interfaces became mainstream, monetisation was inevitable. The interesting question was never whether ads would arrive, but how quickly platforms could introduce them without damaging trust. OpenAI is clearly trying to thread that needle by separating ads from answers and emphasising user control. Whether users accept that in practice is another matter. Either way, the commercial web is reorganising itself around conversational interfaces faster than many brands still seem to realise.

Also worth watching: ranking volatility stayed hot into March

Even after the Discover core update completed on 27 February 2026, ranking volatility remained unusually heated into early March. That left the industry in the familiar position of seeing movement without getting much formal explanation.