Answer engine optimization · News
A bank just said "Generative Engine Optimization." Here's what that actually means for you.
JPMorgan Chase dropped its 2026 Emerging Technology Trends report this month, the sort of thing aimed at enterprise clients and institutional investors rather than a med spa owner in Austin. Dig into the agentic commerce section and there's a named discipline sitting right there: "Generative Engine Optimization (GEO) becomes critical, with AI-driven agents now translating intent into transactions, frequently bypassing conventional research and comparison. Winning brands will focus on influencing the signals and rules that drive agent decision-making, not just consumer discovery."
I've been making this exact case to business owners for months, backed by our own audit data. Ten independent med spas in Los Angeles, zero named by ChatGPT or Perplexity. We ran fifty wellness businesses across five categories and found two thirds invisible to both models. Those numbers came from real queries, counting what showed up. A bank's technology report is different evidence. It isn't selling anyone an audit. It just confirms the shift we've been measuring at the small business level matches what a major financial institution is telling its enterprise clients to plan around.
What the report is actually about
Read the full document and the local business angle barely registers. JPMorgan's report covers four themes: data architecture behind AI models mattering more than the models themselves, infrastructure buildout (global AI capex hit $400 billion in 2025, projected past $600 billion this year), agentic commerce, and synthetic user simulation for testing and security. The GEO line lives inside the agentic commerce section, which is mostly enterprise data governance and how companies grant and revoke permissions to AI agents acting on their behalf.
Worth stating plainly: a report written for corporate technology leaders isn't a study of your industry. It's a signal about where the buying behavior it depends on is headed. If agents are increasingly the ones "translating intent into transactions," the question of which sources those agents trust stops being a marketing nice-to-have and becomes infrastructure, the same way a business's phone number showing up correctly on Google Maps became infrastructure a decade ago.
The number worth keeping, and the one worth dropping
The report puts a size on where this is headed. The AI browser market goes from $4.5 billion in 2024 to a projected $76.8 billion by 2034, compound annual growth rate above 32%. That's JPMorgan's own figure, drawn from their cited research.
The 42% stat that's been circulating, the one claiming 42% of enterprise leaders are already testing AI-native browsers, is also in the report, but with an attribution most retellings drop: JPMorgan is quoting a separate Greyhound CIO Pulse survey rather than reporting its own finding. Worth knowing if you plan to cite it yourself.
Then there's a stat I couldn't verify at all. A secondhand writeup of this report claimed companies with systemic AI integration saw 1.7 times the revenue growth and 3.6 times the shareholder returns of companies still experimenting. I went back to the source PDF specifically to find that line. It isn't there. Somewhere between the bank's report and the summary that reached me, someone added a number that sounds exactly like the kind of thing a report like this would say, and it very nearly went into a client-facing document before I checked.
That's the same failure mode we've documented on this blog before with vendor benchmarks that get repeated until nobody remembers where the original figure came from. A stat traveling through two or three retellings without anyone opening the primary source is how a business ends up making decisions on a number that was never real. The fix here was the same fix we tell clients to apply to any AI visibility claim: open the actual document, find the actual sentence, drop what you can't confirm. In this case that meant losing what would have been the two most quotable numbers in the piece, in exchange for what's left being true.
Where this connects to what you're actually deciding
Strip away the enterprise framing and the report describes the same mechanism our audits keep finding at the local level. When someone asks ChatGPT for a recommendation, the model isn't ranking a page of links, it's picking a small number of sources to trust and building an answer from them. The report frames that as brands needing to influence "the signals and rules that drive agent decision-making." Our data shows what that looks like in practice for a small business: the businesses ChatGPT names tend to have treatment-specific pages, consistent listings, enough review volume to clear whatever trust threshold the model is applying. The businesses left out have general homepages and thin review histories, whatever else they're doing right.
None of that changes because a bank wrote about the enterprise version of the same problem. What changes is that the argument for taking this seriously no longer rests only on a niche audit firm's own data. A source with no reason to hype AI marketing spent eleven pages telling its institutional clients that the rules for getting chosen by an agent are becoming the rules that matter, and used our industry's own term to say it.
If you want to know where your own business stands against that threshold today rather than waiting for it to matter more, that's exactly what the free check runs.
Sources: JPMorgan Chase, 2026 Emerging Technology Trends report; Greyhound CIO Pulse 2025 (as cited within the JPMorgan report).