Lily Ray

Big news! Google just put out its first official article on optimizing for AI search (AEO/GEO)!

TL;DR:

⭐️ SEO is still the foundation for AI search.

⭐️ Google sees optimizing for AI search as SEO, so they’re still calling it SEO.

⭐️ Create non-commodity, "people-first" content.

⭐️ Ignore most "GEO/AEO hacks like chunking and llms.txt.

They also put out another document today that spam policies in search apply to AI Overviews/AI Mode 👀👀

Link to both in comments.

It’s happening: Google just officially announced that Gemini will power Apple Intelligence and Siri.

Is this game over? Cuz it feels like game over.

Link to tweet in comments.

The worst thing you can do for your AI search visibility is destroy your SEO.

I've shared dozens of examples throughout the year of sites quickly losing SEO visibility on Google, with ChatGPT citations following the same trajectory shortly after. I've added a screenshot from my upcoming talks with a couple of examples of this, as well as a similar slide by Natzir Turrado, who recently shared the same thing. (If you don't read Spanish, it says "if you fall in Google, you fall in AI.")

Yet many of the things I've seen folks recommending for GEO are bad ideas for SEO.

I believe a lot of this stems from new GEO practitioners not having enough SEO experience to make good long-term decisions. SEO isn't something you can learn overnight. You need years of experience to understand the cyclical nature of how once-effective tactics can actually become liabilities once search engines start identifying and treating them as spam.

This is especially true for the "too good to be true" tactics that get popularized and shared throughout the industry. I like to assume that the folks on the search engines' spam teams are always watching what we're up to.

If you're comfortable with rapid growth for a few months and then crashing on the other side, go for it. But most brands can't take that kind of risk.

Years of SEO experience give you the context to anticipate when a too-good-to-be-true tactic is really just a loophole, and to play "defensive" SEO. SEO is as much about what NOT to do as it is about what to do.

We've been in a bit of a gold rush with AI search for the past two years, where lots of new approaches have truly worked to manipulate AI responses and citations for a short period. But those tactics are clearly catching up to many sites now. I've been documenting dozens of examples of sites algorithmically crashing or receiving manual actions this year for excessive SEO/GEO-driven spam. Plus, we now have Google employees publicly discouraging "inauthentic mentions" and other GEO hacks, so we know it's on their radar.

I'm also working on new research showing that LLMs like ChatGPT and Claude are clearly taking steps to mitigate spam and manipulation in their responses and citations. Stay tuned for that!

In my opinion, the best approach is the one that leaves your brand being cited and recommended *after* the AI companies clean up the spam.

Google recently reminded everyone that "non-commodity content" is what wins in search, meaning content with a unique perspective, real E-E-A-T, and genuine lived experience. (BTW, this is an extension of what Google has been promoting in its quality guidelines for years.)

I made a half-joke the next day that many marketers' first thought would be: "How can I automate non-commodity content with AI?"

Real talk: the entire point of non-commodity content is that it requires what AI alone can't deliver: real human experience, real expertise, real mistakes, real product testing, real opinions, real emotions, etc. These are the kinds of insights that only come from actually living through something, and the valuable "human" stuff that other humans actually want to read.

Google's Search Quality Rater Guidelines made a change a couple of years back to emphasize the amount of "effort" that goes into building a page. While content creation and scaling have become dramatically easier with AI, I don't see Google's fundamental approach changing much. If anything, this leveling of the technological playing field raises the bar for what truly helpful, original content looks like.

If a piece of content can be built with a few button clicks and a 25-minute Claude session, there isn't a lot of effort or originality behind it, even with the best possible workflows and a polished-looking deliverable.

Take product review sites as an example - as this is a category where Google has become extremely strict when it comes to evaluating content quality. The better-performing product review sites actually buy the products, test them for weeks or months, take their own photos, film side-by-side comparison videos, and document genuine pros and cons from hands-on use. A single piece of content like that can take *days or even weeks* to produce. That's real expertise, real experience, and real effort, and it's exactly the stuff that AI can't replicate.

What will stand the test of time, and what I believe search engines and AI assistants will continue to promote, is the hard work being done by sites with real humans providing their opinions, their experiences, their humor and authenticity, and years of detailed expertise. The kind of evaluations only a human can do.

If there's a middle ground where AI assists the process and creates efficiencies, but the content is still truly evaluated and generated by genuine human expertise, then *that's* the happy medium worth chasing.

But my concern is that - in the spirit of AI - many companies will try to automate something that fundamentally shouldn't be automated. They'll build "non-commodity content" workflows and miss the entire point of why users prefer it.

The sites that succeed long term are the ones with real human evaluation happening throughout the process. That part can't be automated. By design, it takes time and effort. That's what makes it good content in the first place.

This year's summer speaking tour has officially come to an end, and it was a blast!

Thank you Athens SEO, Croatia SEO Summit, CAMPIXX Berlin, Peec AI, and Search 'n Stuff London for all the awesome conferences & meetups.

My topic this year was "How SEO & GEO Work Together - Do Great GEO Without Destroying Your SEO" and I spoke about some of the common approaches sites have been using to try to win in AI search that are actually backfiring and hurting their organic performance. Then, when you hurt your organic performance, your performance in AI search follows suit (since language models much of their information from top-performing search results).

I also included a section about what things are worth focusing on and how I'm approaching AEO/GEO.

Just uploaded my deck to Slideshare - the link is in the comments!

Pro tip: just because your brand is being recommended as #1 on your personal, logged-in ChatGPT account certainly does not mean others are seeing the same results.

Over the years, I’ve used various tools, Chrome extensions, and (obviously) the browser itself to diagnose JavaScript-related SEO issues. But now with AI search, I’ve found myself needing a quick and easy way to check if a page’s JS setup is “LLM-friendly” - so I built a tool myself.

My “LLM Content Visibility Scanner” shows you what (non-Google) AI crawlers like GPTBot (ChatGPT), ClaudeBot and PerplexityBot actually see when they look at your page, without executing JavaScript (the way most LLM bots fetch content.)

Just paste a URL and it quickly audits the raw HTML for client-side rendering issues, missing structured data, paywalls, alt text gaps, and crawlability problems.

It then hands you a prioritized mini-action plan with specific URLs to address, plus some other metadata information.

Note: there are some known limitations listed at the bottom, so the tool isn’t always 100% perfect. And it’s certainly not meant to serve as a replacement for a proper technical SEO audit. But it’s a good starting point and a quick tool to use for this purpose when you’re on the go.

Let me know what you think and feel free to DM me any feedback/bugs you might encounter!

https://lnkd.in/gruiC332

It seems like ChatGPT was heavily reliant on Google's search results for web retrieval up until a few months ago, and now it looks like OpenAI is leaning on Bing results more heavily (again), while building its own cache/index.

The timing makes sense: Google sued SerpApi in December, the scraping company that reportedly supplied OpenAI with Google SERP data after Google declined to give them direct index access. Between that lawsuit, Google's 'SearchGuard' anti-scraping crackdown, and OpenAI building its own index, it makes sense that OpenAI would try to decouple from Google right now.

There's also some direct evidence for this: a new "bing" result_source value started showing up in ChatGPT's network traffic in late June, rolling out to some accounts but not others (h/t David Konitzny and Suganthan Mohanadasan, who have been digging into this).

Are others seeing similar things in their data? Does anyone know more about this? And if so, can we chat? 😇

I've said this before, but it's still true: when Britney posts, I listen closely.

"Stop optimizing for 'AI.' Optimize for search engines (so retrieval-based AI can cite you) + earn third-party coverage (so the model already knows you before the prompt is typed)."

🫡

Pro-tip: with ChatGPT adding "official" to many more fan-out queries, it's a good time to consider whether adding "Official" makes sense to things like official brand homepage