Healthy skepticism. A new acronym appears, agencies attach retainers to it, and everyone claims it is the future. So let us separate what is genuinely new about generative engine optimization from what is recycled SEO with a fresh invoice.
The behavior shift is measurable and large: people ask ChatGPT, Gemini, Claude, and Perplexity for recommendations and act on the answer without ever seeing a results page. Those answers name a handful of brands, not ten blue links, so being excluded is invisible and expensive. That part is not hype, and you can verify it in an afternoon: ask the assistants your buyers questions and see who they name. The free does exactly this.
Much of the work sold as GEO is good content and technical hygiene that SEO always rewarded: direct answers, structured data, consistent facts, real citations. If someone sells you GEO as magic separate from that foundation, walk away. The honest picture is in GEO vs SEO: same inputs, new judge. What genuinely is new is the judge: assistants weigh conversational context and third party trust differently than a rankings algorithm, and they decide upstream of keywords, as we argue in what you are being taught about AI visibility is wrong.
Skip the think pieces and look at three observable facts. First, assistant usage: hundreds of millions of people now ask AI for recommendations weekly, and those answers name three to five brands where a search page showed dozens. Second, the referral pattern: brands increasingly see buyers arrive already decided, mentioning the assistant that sent them, traffic that shows up as direct or branded search in analytics. Third, the vendor behavior: every major SEO platform has shipped AI visibility features in the last two years, which is what incumbents do when a category is real. None of this requires believing anyone's marketing, including ours; it is all checkable from your own data and your own assistant sessions this afternoon.
What remains legitimately uncertain is magnitude per category. A B2B firm whose buyers research heavily in AI conversations has a very different exposure than a local shop found through maps. That is why the honest starting move is measurement, not spend: find out what fraction of your category's moments already produce brand recommendations, and let that number size your investment.
Do not argue the acronym; demonstrate the shelf. Open the four assistants in a meeting, ask each one two real buyer questions, and let the room watch who gets named. If competitors appear and you do not, the conversation changes from is this real to what does it cost to fix, and you can bring the published numbers from our pricing page next to any agency quotes. If nobody gets named, you have found an unclaimed shelf and the argument becomes about speed. Either way, put the result next to one more number: what a single customer is worth in your business. Shortlists of three to five, times your close rate, times customer value, is the size of the prize being reshuffled. The scoreboard for tracking it over time is in how to know if your GEO is working.
Two skeptical claims hold up under inspection: attribution really is murkier than paid search ever was, AI-referred buyers arrive with no clickstream, and vendor hype really has outrun vendor evidence across this category. Take both seriously and they sharpen your program: murky attribution argues for frozen baskets and trend measurement rather than giving up on measurement, and hype argues for published pricing and verifiable baselines rather than abstinence. Where skepticism turns expensive is the leap from the vendors oversell to the behavior is not happening, because the behavior is checkable in an afternoon and compounds while you wait. The skeptics who ran the afternoon test and started quietly are in the strange position of being both right about the hype and early to the shelf, which is the best seat in any new channel.
The inputs overlap heavily, but the judge changed. Rankings algorithms order pages; assistants recommend a shortlist of brands based on context and verifiable trust. Optimizing for a shortlist is a different game than optimizing for position.
Ask the four assistants the questions your buyers ask. If competitors are named and you are not, that is demand you are losing invisibly. If you are named everywhere, GEO is maintenance for you, not an emergency.
The mentions assistants trust take time to build, which means early movers compound. Waiting until AI referrals show up in your analytics means starting after competitors are already the default answer.
The acronym might. The behavior, people taking recommendations from AI conversations instead of search pages, has grown every quarter since assistants launched and underpins the business models of the largest AI companies. Bet on the behavior, not the term.
Partly justified annoyance at grifters selling old tactics under a new name, and partly turf. The legitimate critique, that the inputs overlap heavily with good SEO, is true, and it does not change the fact that the judge and the shelf changed.
One afternoon: run the free GEO Grader, ask the assistants ten real buyer questions manually, and count who gets named. Total cost is zero and you will know your exposure by dinner.
Two clean quarters of baseline showing your category's moments produce no brand recommendations, plus no drift in how buyers describe finding you. Test it; do not assume it.
Assistant usage numbers come from the platforms and standard traffic trackers; your own category's exposure you measure yourself in an afternoon, which beats any industry survey for deciding your budget.