For the first time in nearly four years of generative AI disrupting the search landscape, Google has finally published some official guidance on how to optimise for AI search. The documentation, entitled Optimizing your website for generative AI features on Google Search funnily enough, is genuinely useful, and at 22i we very much welcome it. But we’d be doing you a disservice if we didn’t offer a genuine, balanced, and informed reading of it rather than simply sharing it without any critical thinking.
Why Guidance Matters and Why it Took So Long
One of the biggest frustrations of the AI search era has been the lack of any real authoritative framework. When Google stays silent on a subject, the search industry, naturally, quickly moves to fill the void. And that, to be frank, is not always the best thing to happen. We’ve seen a surge of low-quality AI-generated content (AI slop), schemes to game AI Overviews, and entire consultancy offerings built on the speculative “hacks”. but with little evidence behind them.
Google has, historically, been quicker to tell us what not to do than to actually show us the way forward. Its scaled content abuse updates eventually dealt with the worst of the AI content flood, but the lag there was for over a year after seasoned professional copywriters had to endure millions of inexperienced non-writers simply pressing buttons. Too little too late. However, in the vacuum created by the guiding voice staying muted, practitioners across the SEO industry, ourselves included, have been testing, monitoring, and building workflows based on the best available evidence that we’ve had.
That testing hasn’t been wasted. Much of what the search industry has developed independently aligns closely with what Google has now confirmed. But confirmation still matters, and this official guidance provides it. However, it’s always wise to take some of the official Google line with a healthy level of skepticism.
The Central Message: Strong SEO Remains the Foundation
The headline finding from Google’s guide is crystal clear: foundational SEO best practice is still the most effective strategy for visibility in generative AI search. Google puts its guidance here very directly; businesses should “apply foundational SEO best practices to generative AI search features.”
That’s decades of established discipline validated in a single sentence. Content quality, technical health, structured architecture, page experience, crawlability, and relevance signals all remain central to AI search success. If your SEO foundations are strong, your position in AI-powered search results is firmed up alongside your position in the traditional SERPs. So that’s a good start.
On content specifically, Google states that creating material people find unique, compelling, and useful will, in the long run, have more influence on AI search presence than any other single factor. The key attributes they highlight are:
- original point of view,
- helpful and reliable information,
- logical organisation,
- quality media,
- user-focused intent,
- and adherence to spam policies.
These are the same principles that have defined excellence in digital marketing for so many years. No change there then.
And well-structured pages matter too. Semantic HTML, sound JavaScript SEO practice, good page experience, minimal duplication, and strong crawl hygiene all feature in the guidance; all of these fit firmly within the scope of comprehensive SEO mastery.
A Note on Trust: Reading Google’s Guidance Critically
However, all that said, we’d be a bit remiss not to acknowledge the context in which this latest official Google guidance arrives (And just ahead of next week’s Google I/O developer conference). If you cast your mind back to 2024, there was an internal data leak at Google that revealed they’d been using ranking signals they had publicly denied; User engagement metrics, Chrome browsing data, and algorithmic micro-adjustments known as “Twiddlers” were just some of the previously denied evidence found in the leak. So the gap between Google’s public statements and internal practice has form!
But that doesn’t make this new guidance wrong. Much of it is corroborated by independent evidence and aligns with how generative AI systems demonstrably process and surface content. But, it does mean that we read it with professional discernment rather than uncritical acceptance, which is exactly what you’d expect from an experienced agency partner like 22i Design & Digital.
GEO, AEO, and the Terminology Landscape
Google’s AI search guide then makes an interesting editorial choice: in a 2,400-word document, the term AI search appears 21 times, while GEO (Generative Engine Optimisation) and AEO (Answer Engine Optimisation) are each mentioned only four times, and largely to contextualise rather than endorse them.
That is a very deliberate signal from the Mountain View behemoth. Google frames GEO and AEO as terms “common online” that have generated misconceptions alongside genuine insights. It’s also worth noting here that Microsoft’s Bing, along with much of the broader AI search ecosystem including Perplexity and other generative platforms, actively uses the GEO moniker. The terminology reflects real differences in how these platforms index and surface content, and a sound AI search strategy acknowledges the full ecosystem, not just Google’s own properties.
Here at 22i, our approach has always been to focus on solid, platform-agnostic quality signals rather than chasing single-platform definitions. So optimising for genuine authority, topical depth, and entity clarity serves your visibility across Google AI Mode, AI Overviews, Bing Copilot, ChatGPT, and Claude alike.
LLMs.txt, Schema, and the “Myths”
The next section of Google’s guide that has generated the most industry discussion addresses a range of practices that have proliferated in the absence of any official direction, that void we spoke about earlier. Google characterises several of them as misconceptions, so it’s worth unpacking each with forensic honesty.
- LLMs.txt: Google is unambiguous that this file has no bearing on how its systems index or surface content. Research has also suggested it has little measurable effect on AI search inclusion more broadly. That said, implementing an LLMs.txt file is a minimal investment, a matter of minutes each month to maintain, and other AI platforms and crawlers may yet assign it value. We continue to recommend it as a low-effort, no-downside hygiene measure, while being clear that it is not a meaningful ranking lever on its own. We’d say here that we;ve heard ChatGPT results may be pinched by LLMs.txt, but a little bird tells us Claude is OK with the unofficial protocol.
- Structured data and schema markup: Google’s position here is importantly nuanced: they are not saying abandon schema. They are saying don’t over-index on it as an AI search shortcut. Schema markup still remains a meaningful signal for the semantic web, aids machine readability across multiple platforms and use cases, and supports entity disambiguation, which is directly relevant to how AI systems build topical understanding. FAQ schema has been formally retired by Google for SERP features (allegedly), but structured data, more broadly, is alive, well, and worth maintaining thoughtfully.
- Content chunking: Writing in well-organised, logically segmented sections that are easy to read and extract information from is simply good copywriting. If you write clearly and structure your content for human comprehension, the “chunking” principles follow naturally. Writing in fragments specifically for AI ingestion, on the other hand, produces stilted content and doesn’t serve your readers, which ultimately doesn’t serve your rankings either. Think about it this way – the “chunking” is really about placing concepts into sizeable paragraphs, and that, dear readers, is what good writers do anyway, am I right?
- Rewriting content for AI systems and seeking inauthentic mentions: these are rightly called out as counterproductive. Content that’s been distorted for machine consumption tends to be worse for human readers, and artificial mention-seeking carries the same risks as any other manipulative link-building strategy. We’ve seen clients attempt to rewrite solid content purely because they’ve read that this is the new thing, and we’ve had to stop them in their tracks. Nothing can be worse than an obviously ChatGPT-generated article that so obviously looks like it had no effort. And that, kids, is another signal – “low effort”. Please put effort into your work and Google (and others) should hopefully recognise and reward it.
What This All Means for Your Business’ Website
The practical upshot of Google’s guidance, read with appropriate context, is that businesses with a strong, well-maintained digital presence, are well positioned for this era of AI search. The fundamentals haven’t changed but the surface they operate across really has expanded.
For 22i’s clients, this translates into the same strategic priorities we’ve always advocated:
- authoritative, original content that clearly communicates expertise;
- technically sound websites that are easy to crawl, index, and understand;
- structured data applied with purpose, and
- a consistent focus on topical depth and entity clarity rather than just keyword volume.
The world of AI search is (still) evolving rapidly. And according to Similarweb data, ChatGPT’s share of AI search traffic has fallen from 77.6% to 53.7% over the past twelve months, while Gemini has grown from 7.3% to 26.7%, and other platforms are gaining ground too! (Nice work, Claude). Google I/O 2026 is imminent. The signals will keep shifting, but the foundations won’t.
If you’d like to understand how your current digital presence performs against these criteria, or want to build a strategy for visibility in both traditional and AI-powered search, we’d welcome the conversation. Call us on 01252 692 765 or leave a message.
