TLDR, GEO gets you mentioned. AX determines what happens next.
Story time! I recorded a podcast last week with Anne Bertucio de Hollembeak, COO of ROOST, (forthcoming, watch your MonkCast feed) and we found ourselves talking about what seemed to be the same thing but with different words. She said GEO and I said AX. As a resident of the Pacific Northwest, GEO reminded Anne of a geoduck (which I didn’t use as the header image, so, you’re welcome everyone), while I have been focused on AX, so we mutually agreed to use AX. However, neither Anne nor I felt confident in the nuances of the distinction, so I dug a little deeper.
As the more developer-focused concept, AX has been a big concern for us at RedMonk. More and more of our conversations have touched on the idea, and our clients have increasingly reached out to us to discuss how to do AX well. While our MonkCast transcripts certainly qualify as GEO, I wanted to discuss specifics about where AX meaningfully differs from GEO.
So here’s my breakdown targeted to practitioners in the software space. I focus on what each term means, a bit about their history, and why both matter today.
What is Generative Engine Optimization (GEO)?
Sometimes called AI SEO, Answer-Engine Optimization (AEO), or LLMGEO, GEO emerged from the SEO and marketing industry’s realization that as search moves from Google to Chatbots, discoverability requires a technological update.
GEO first appeared in November 2023 in “GEO: Generative Engine Optimization,” a paper published to arXiv which was later peer reviewed and presented at KDD 2024. The authors argue that content creators have “little to no control” over when and how their words show up, so the team built a benchmark, tested content tweaks, and reported visibility gains of up to 40 percent inside generative engine responses. The tactics that scored well include adding statistics, quoting named experts, and citing your sources.
By 2025 the term had escaped the lab. New York magazine was running the headline “SEO Is Dead. Say Hello to GEO,“ and an entire industry of consultants and marketers willing to help vendors improve their GEO has popped up. Why the interest?
GEO concerns itself with traffic, and traffic is newly borked. Pew Research tracked the browsing of 900 US adults and found that when a Google AI summary appeared, people clicked a normal search result in 8 percent of visits, against 15 percent when no summary appeared. Links inside the summary itself got clicked in 1 percent of visits. While Google has called the methodology flawed, the problem of conversion is real and has serious financial repercussions.
What is Agent Experience (AX)?
In January 2025, Netlify co-founder and CEO Mathias Biilmann published a post defining AX as:
the holistic experience AI agents will have as the user of a product or platform.
He put it in a lineage: Don Norman named user experience in 1993, and Jeremiah Lee named developer experience in 2011. Similar motives, new user, and that new user is a computer. The reason Biilmann argues AX is needed is because companies are bolting shallow AI features onto their products instead of asking the harder question: when your customer sends an agent to do a job inside your software, can it?
But overlap between AX and GEO is real, and Biilmann acknowledges it himself in the follow-up post to the one coining the phrase. Of his four areas of agent experience he identifies, Access, Context, Tools and Orchestration, he files GEO beneath Access on the grounds that the agent has to know your product exists before anything else matters. Correct, and mildly humbling to many in the developer crowd, who would rather not think about marketing at all.
AX v. GEO: Quick, Overly-Reductive Differences
- A marketing lead might say “we need a GEO strategy.” A platform engineer could say “our AX is broken.”
- GEO is upstream of the introduction. AX is everything after it.
- GEO is measured in citations, mentions and share of answer. AX is measured in whether the task finished.
- If you sell to people, GEO is your problem. If you sell something an agent has to operate, AX is your problem.
LLMs.txt
There’s been a lot of chatter about the contested usefulness of llms.txt, a proposed file you drop at your root to helpfully point models at your best pages. Many folks I speak to refer to this standard as the best way to address the post-search engine discoverability problem. Unfortunately, that is not currently the case.
A bit of history. In 2024, Jeremy Howard, co-founder of Answer.AI, proposed LLMs.txt as a standard for AI website content crawling. Since then, and despite extensive tech media coverage, adoption has been slow. In June 2025, for instance, John Mueller, Senior Search Analyst at Google, posted to Bluesky “FWIW no AI system currently uses llms.txt.” A year later, Evil Martians measured two months of their own traffic and concluded that Mueller was “basically right”: the file gets fetched by crawlers, not by the agents it was written for.
The Future of AX & GEO
A lot of the AX/ GEO conversation comes down to persona and use case. AX and GEO are two disciplines that arrive at very similar advice, but from different places. The question today is partly one of legitimacy. Snake oil was rampant in the early days of SEO. Blog posts and consultancies touted checklists and incantations guaranteed to improve ranking. This heritage still feels strong.
Although there are definitely things that can help (MCP can make sure an LLM accesses the most up to date docs, etc), no single “slap on an MCP server” panacea is going to solve a vendor’s discoverability problems or problems tied to a devtool’s appropriate use. In short, there remains much to be skeptical about when it comes to GEO and AX solutions. Recommendations on how to make improvements around each run the gamut from sensible to occult.
From my perspective, what’s worth knowing is that the problem of empowering agents to recommend and use third party products, and particularly devtools, is currently in flux and agents require both AX and GEO to be successful.
Disclosure: Google is a RedMonk client.