Episode Description
Patrick Stox spent six and a half years inside Ahrefs before going independent, and his position on GEO is not the one the acronym wars want. SEO is a subset of it, not a synonym for it, because the fastest way to change what a machine says about your company is to fix the thing people are complaining about. The rest of the conversation is what an operator does once that is true.
Key takeaways- Influencing AI answers is a product problem first. If the internet is complaining, the model repeats the complaint.
- Retrieved and cited are two separate contests. Being pulled into the candidate set is table stakes; getting quoted is a second bar you have to clear.
- The frontier past AI content at scale is information gain. His method is to interview the experts already on your payroll, two or three times a week.
- Do not run a programmatic project without unique data. A derived metric nobody else has computed, not a bigger pile of the same rows.
- AI search systems have not decided to forgive anything yet. Google chose to forgive link spam. That was policy, not physics, and nobody has made that call for the answer engines.
- Why SEO is a subset of GEO
- The forgiveness nobody has granted yet
- Retrieved is not cited
- Information gain, and the experts you already employ
- Chapters and timestamps
- People, ideas and sources mentioned
- Questions this episode answers
- Go deeper
If you’ve got a problem with your product, the easiest way to get the LLM or the AI search system to say something different? Go fix that. Because if people are complaining on the internet, that’s what it’s gonna say. So it’s business, it’s branding, it’s so much more.
— Patrick Stox
The claim is not that on-page work stopped mattering. It is that the ceiling on it is lower than the industry admits, because the systems are summarising a reputation you do not own. Stox draws a scale line through the middle of that: a local business can put an SEO on sentiment and move it. A large company cannot staff its way to changing the opinion of the internet.
Which makes GEO a business, brand and product remit that happens to include search work, rather than a rebrand of the search work.
The forgiveness nobody has granted yetSEOs got lucky with Google. We got Penguin, and for a few years that was really painful — and then they’re like, yeah, we forgive all your bad links. These AI search systems may make a different decision. How long will they hold a grudge against you?
— Patrick Stox
Google decided, around 2016, to stop holding historic link spam against sites. That decision could have gone the other way and stayed that way for another decade. Nobody at OpenAI, Anthropic or Perplexity has made the equivalent call, publicly or otherwise.
The practical reading: tactics that work today carry an unpriced liability. That is not an argument against moving fast. It is an argument for knowing which of your tactics you would be embarrassed to have logged.
Retrieved is not citedYou get a bunch of pages, might be 90 pages retrieved, and then 13 of them get cited... Well, why were some retrieved but not cited? That’s a really good question to look into. And it’s usually like, you were relevant enough, but you didn’t quite say the right thing, or in the right way.
— Patrick Stox
The sharpest operational idea in the episode, and the one most reporting misses entirely. A page that gets retrieved and dropped is already close enough to win. Stox calls that gap the low-hanging fruit of the AI era, because you are not fighting for relevance any more, you are fighting over phrasing and specificity.
How Google finds things at all
Most people will tell you sitemaps, or crawling. Well, there’s more. RSS feeds are a thing. Pinging the sites — the WebSub, PubSubHubbub, whatever... Submissions in GSC, the indexing API.
— Patrick Stox
Worth keeping when you are diagnosing a discovery problem, because most audits check exactly two of those inputs and stop.
Information gain, and the experts you already employAnyone can put out any number of pages. But if you’re just using it for content — just the content — you’re not providing resources, you’re not sourcing where you got something. So anyone can do it, but it’s a matter of: can you do it well?
— Patrick Stox
His alternative is unglamorous and almost nobody runs it. Message five genuine experts inside your own company, two or three times a week, over Slack or email or an AI-run interview, and build from what they say. The raw material is already on the payroll.
Generally I wouldn’t even recommend a programmatic project unless you’ve got something unique, different... Everyone’s got reviews, but what if you did scoring? What if you did a market analysis? What’s the sentiment? What is the brand search volume for each of these companies?
— Patrick Stox
And on the Helpful Content Update, he is blunt that prose is not the moat:
ChaptersI don’t think writing will get you but so far. Part of the thing they mentioned with that update was images — images and videos that show you’ve got hands-on with the product you’re talking about... Not you pulled the stock photo offline from the manufacturer and then wrote a bunch of stuff.
— Patrick Stox
| Time | What happens |
|---|---|
| 00:00 | Introduction and guest background |
| 00:49 | From developer to SEO, and six years inside Ahrefs |
| 02:27 | Dream tools and infrastructure projects |
| 09:30 | Tech SEO Connect, and what a conference is actually for |
| 11:21 | Content at scale, and what comes after it |
| 14:01 | Leveraging unique data instead of volume |
| 15:00 | The hot take: SEO is part of GEO |
| 17:48 | Past algorithm changes, and the risk nobody has priced |
| 29:00 | Backlinks, and their changing role under AI |
| 33:05 | Google's infrastructure, data sources and internal systems |
| 35:00 | Retrieval, ranking and citation |
| 36:36 | Expertise and content-quality cues |
| Entity | What it is |
|---|---|
| Patrick Stox | Independent SEO consultant. Six and a half years at Ahrefs, four at IBM, started as a developer |
| Tech SEO Connect | The technical SEO conference he co-runs, in Durham, North Carolina |
| Ahrefs | Where he worked on Page Inspect, site structure, Site Audit and the v2 to v3 infrastructure change |
| Helpful Content Update | The Google update discussed at length, including the added Experience in E-E-A-T |
| Cats.com | His standard for hands-on expertise: a site that sends pet food out for laboratory testing |
| Information gain | The through-line: content that adds something the corpus does not already contain |
- Is GEO just SEO with a new name?
- Why do some pages get retrieved by an AI search system but never cited?
- Will AI search systems forgive spam the way Google forgave link spam?
- What comes after publishing AI-generated content at scale?
- When is a programmatic SEO project actually justified?
- How does Google discover content beyond sitemaps and crawling?
- What did the Helpful Content Update actually reward?
Every link above goes somewhere different. These are the ones not already mentioned above.
- the consensus machinery underneath that retrieval gap, where Malte Landwehr shows how answer engines resolve a question before they pick who to quote.
- how a complaint reaches the answer before you do, which is Ann Smarty tracing the retrieval path Stox is telling you to cut off at the source.
- this episode as a plain-text file for AI assistants, carrying the full cleaned transcript with a speaker label on every turn. Paste it into ChatGPT or Claude and interrogate it directly.
patrickstox.com, and Tech SEO Connect at techseoconnect.com.
Cite this episodeStox, Patrick. Interviewed by Jeremy Rivera. “Patrick Stox on Building in the GEO Era.” The Unscripted SEO Podcast, 26 June 2026. https://unscriptedseo.com/patrick-stox-on-building-in-the-geo-era/