Big tech just said the quiet bit out loud: GEO builds on good SEO, and good SEO still needs originality

Google has published its clearest advice yet on how to show up in AI Overviews and AI Mode. Meanwhile, LinkedIn has AI slop in its sights. It’s a timely reminder for B2B marketers: AI is changing how research is delegated, not eliminating buyer intent or the human evaluation moment. Make useful, original content for humans, make it technically accessible, and stop trying to autospam the open internet.

Back in 2024, I wrote that AI would kill content as we know it, and that this was probably fine. Plausibility is not trust. Once anything can generate a passable page in seconds, the reader’s first question changes from “Is this useful?” to “Is any of this true?”

Earlier this year, I wrote about the significance of the meatspace: the physical human layer AI systems still cannot fully fake. Interviews. Fieldwork. Customer conversations. Original data. Something that exists because someone with arms, a mouth and one of those spinal column things spoke to another sentient bag of flesh and came back with something an LLM couldn’t find on the internet.

Google’s brand new guidance that dropped last week lands in that same annoying patch of reality. So does the post-I/O panic about whether “Search is dead”. Google is not replacing search so much as adding an agent layer on top of it. Users still have intent; they are increasingly delegating the research phase to AI.

Google is not asking for GEO hacks

On 15 May 2026, Google Search Central published a new resource on optimising for generative AI experiences in Search. Its message is simple: optimising for generative AI search is still optimising for the search experience, and therefore still built on good SEO.

That matters because a small industry has tried to turn AI visibility into a separate priesthood. AEO. GEO. LLMO. Add an llms.txt file. Mention yourself on Reddit 400 times under a blood moon. 

There absolutely are new mechanics to understand. AI Mode uses query fan-out, running multiple related searches from one query and stitching together a broader answer. Google is also surfacing public discussions, social posts and first-hand perspectives because people want advice from people, not beige averaged web soup.

But the instruction is not ‘spray more AI copy into the pipes’. Google’s own summary mentions valuable, unique, non-commodity content, mythbusting AEO and GEO misconceptions, and why SEO best practice remains foundational.

This is not Google being subtle. It is Google making it really clear that the rise of GEO is not a licence to autospam the open internet just because you’ve found the Automations button on Codex.

Good GEO builds on good SEO, because B2B buyers reward proof

What Google is furiously telegraphing is this: GEO is not replacing SEO. It’s another layer on top of SEO, exposing which bits were always doing the real work.

Technical accessibility still matters because machines need to crawl and understand your site. Structure still matters because everyone appreciates content that gets to the point. Authority still matters because AI systems, like biological journalists and buying committees, look for corroboration. For a B2B buyer, that corroboration is rarely one page. It is a trail of analyst validation, customer proof, product clarity, executive expertise and earned authority that makes a vendor feel safe enough to shortlist.

This is why the PR, SEO and GEO conversation matters. In our earlier piece on the rise of GEO, we argued that brands are surfaced by AI because they are credible, trusted and authoritative across earned and owned channels. Your company FAQ page and earned media profile are now connected. The executive byline, analyst quote, product page, customer story, forum discussion and journalist’s article all become part of the same set of signals.

Less poetically: the machine is trying to work out whether the story holds together. So is the buyer.

That is also the point of our methodology, VISTA. Visibility is no longer just rankings. It is whether a brand shows up consistently across the places AI systems, buyers, journalists and customers use to make sense of a category. Identification, Signals, Trust and Attribution make a brand retrievable, believable and measurable.

We built out VISTA because discussing GEO like it’s a zero sum traffic problem misses the point. An IT director evaluating a fintech vendor through Perplexity, Gemini or AI Mode is not behaving like someone Googling running shoes. They are trying to reduce risk, compare claims, understand integrations, find proof and work out which vendors are credible enough to bring into a human buying process. That means the answer layer is not the end of the journey; it is the research layer that shapes who makes the longlist.

Original content is not a Skill you can download

Fortunately for good marketers, and less so for procurement, here is where the automation fantasy runs out of road.

Yes, AI research at scale is unbelievably effective. It can map search results, cluster themes, compare citations, analyse competitor visibility, summarise expert debate and generally do in minutes what used to require a person, a spreadsheet and maybe some burnout. It certainly can surface connections in data that may lead to new insights that a human may have spotted. This is all valuable.

But research at scale is not the same thing as originality at scale. Original content requires something new to enter the system. It requires human intellect and judgement, yes, but also the tedious physical apparatus attached to them. Someone has to spot the useful question, decide the easy answer is too easy, and get the data, quote, observation, client insight, customer complaint, sales objection, expert caveat or boring operational truth that does not already exist online for any old LLM to feast on in training.

That is the work. In complex B2B categories, it is also the difference between being mentioned and being trusted.

AI can help you find the gap, understand the category and turn a messy interview into something readable. But it cannot create first-hand experience from nowhere. It cannot, despite valiant efforts by the LinkedIn automation community, have a thought on your behalf that is meaningfully yours.

Please do not autospam the internet

There is a reason Google keeps coming back to helpfulness, originality, page experience, duplication and spam, and LinkedIn is now reducing recommendation reach for generic, recycled, engagement-baity AI slop. Different platform, same message: AI assistance is welcome; synthetic thought leadership slurry is not.

If your AI and GEO approach is mostly a way to publish more pages with fewer people involved, you may get a short-term output graph that looks exciting in a quarterly meeting. You may even get some rankings, briefly. But you are building the low-trust content estate users, AI systems and search quality teams are learning to discount.

The better use of automation is not to replace the human layer. It is to protect it. Automate research grunt work so strategists can find the actual point. Automate transcription so interviews with brilliant people happen more often. Automate reporting so the team can talk about meaning, not chart formatting.

The human steps are not cost centres to squeeze. They are value centres to protect. Google’s latest advice just makes the same point in Search Central dialect.

So, yes, do the SEO basics. Make pages crawlable. Structure content clearly. Answer real questions directly. Reduce duplication. Improve page experience. Use schema where it helps, especially when Google or another interface might build a comparison table, calculator or answer module from your information. Track rankings, citations, brand mentions, citation share and sentiment.

But then do the bit that cannot be copied from a checklist: build a point of view, interview brilliant people who know things, commission research, turn customer patterns into evidence, put experts in market conversations, align earned media and owned content, and create work with provenance.

AI helps us see the shape of the problem. It is not a substitute for having something to say, or somewhere to say it.