When there's no results page, there's nothing to rank on.
Somebody asks ChatGPT which company to call and gets three names. No ten blue links, no chance to be scrolled past, no second place. Either a model can read you, verify you and corroborate you somewhere else, or you weren't in the answer.
This is the newest part of my work and the one with the least settled vocabulary. I'll tell you plainly which of it is established practice and which of it is still a bet.
The acronyms, and which ones are worth your time.
This field invented four new terms in about eighteen months and most pages about it will happily sell you all of them. What each one means, and how much of it is settled:
SEO
Search engine optimization
Be a result. The original job: rank on a page of links for a query somebody typed.
Established. Still the largest channel by volume for most businesses.
AEO
Answer engine optimization
Be the answer. Get quoted directly in a featured snippet, an AI Overview or a voice response, where the user never sees a list.
Established enough to plan around. The techniques overlap heavily with good SEO.
GEO
Generative engine optimization
Be the source. Get pulled into what a model generates when somebody asks it for a recommendation.
Newer, less measurable, and worth doing because the work is mostly things you should be doing anyway.
You'll also see HEO used, which depending on who's writing means hybrid engine optimization, human engine optimization or human experience optimization. Three meanings and no measurable search demand behind any of them, so I don't build around it. If somebody quotes you for HEO, ask which one they mean before you sign.
What a model needs before it will say your name
These systems don't rank. They retrieve, then they generate. That distinction changes what the work is, because the thing you're optimizing for is whether you get retrieved and whether the retrieved material supports a confident claim.
It has to be able to read you
Content that only exists after JavaScript runs, text baked into images, key facts trapped in a PDF or behind a form. If it needs work to extract, it usually gets skipped. Plain, server-rendered HTML wins here for boring reasons.
It has to know what you are
Structured data does the heavy lifting. Organization, LocalBusiness, Service, Person, FAQPage and Product markup tell a machine what your pages mean instead of leaving it to infer from prose. This is the highest-return work on the list.
Your facts have to agree with each other
Your site, your Google Business Profile, your directory listings, your LinkedIn and your invoices should all state the same name, the same address, the same service area. Contradictions lower confidence, and a low-confidence answer just names somebody else.
Somebody else has to say it too
A claim only on your own site is an assertion. The same claim on a directory, in a roundup, in a client story or in a publication is corroboration. Third-party agreement is what turns a mention into a recommendation.
It has to answer the actual question
Pages built around a keyword produce content that circles the topic. Pages built around a question produce a paragraph a model can lift verbatim. Answer first, elaborate second, and put the answer near the heading that asked it.
It has to look like a real business ran it
Named authors with real credentials, dates, specifics, cited sources, evidence of first-hand work. Google publishes this as E-E-A-T and the same signals help elsewhere, because they are just the marks of writing by someone who did the thing.
Not a ranking factor. Still the thing that decides it.
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness. It comes from the guidelines Google gives the human raters who evaluate search quality, so there's no E-E-A-T score in the algorithm to optimize. What there is: a set of things those raters were told to look for, which the systems are then tuned to approximate.
Which makes it useful in a way most SEO advice isn't. You can't game it, and you don't need to, because the checklist is mostly "prove a real person with real experience wrote this". That happens to be exactly what makes a page persuasive to a human buyer too, which is a rare case of the incentives lining up.
It matters more for AI search than for regular search, because a generated answer stakes the assistant's credibility on the recommendation. A model has more reason to name a business it can verify than one it can only find.
What that looks like on a page
Experience
First-hand specifics. A project you ran, a number you measured, a mistake you made. Not a summary of what the field says.
Expertise
A named author with credentials that are checkable, and terminology used correctly rather than decoratively.
Authoritativeness
Other people referencing you. Directory listings, roundups, client stories, quotes in somebody else's article.
Trustworthiness
Real contact details, honest pricing language, dates on content, corrections when you get something wrong.
The work itself
Visibility baseline
I ask the assistants the questions your buyers ask and record what comes back: who gets named, which sources get cited, and where you appear or fail to. That is the number we move.
Entity consistency
One canonical description of who you are, applied everywhere. Site, profile, directories, social bios, schema. Machines penalize disagreement more than they penalize silence.
Structured data build-out
Full JSON-LD across the site with an @graph that connects your organization, your services, your locations and you as a person, rather than isolated fragments on individual pages.
Question-shaped content
Pages and FAQ blocks written around the real phrasing of a real question, with the answer in the first two sentences so it can be lifted cleanly.
Corroboration
Getting you into the directories, roundups and listings that these systems already treat as reference material. On local terms this is often faster than trying to outrank them.
Crawlability for AI agents
Server-rendered content, sane robots directives, clean sitemaps, and a deliberate decision about which AI crawlers you allow rather than whatever your host defaulted to.
What nobody can promise you here
This is a young field with a lot of confident salespeople in it, so it's worth being clear about where the ground is soft.
There is no rank to check.
Ask the same assistant the same question twice and you can get two different answers. Measurement here means sampling many prompts over time and watching a trend, not reading a position number.
Nobody outside these companies knows the weighting.
The mechanisms are partly documented and mostly inferred. Anyone describing the exact ranking logic of a generative answer is guessing with more confidence than the evidence supports.
The volume is not there yet for most local businesses.
AI search is growing quickly and it is still a fraction of the queries that hit the map pack. For a Central Florida service business, the profile and review work usually pays sooner. I will tell you if that is your situation.
Most of the fix is not exotic.
Clean markup, consistent facts, real content, third-party corroboration. If a proposal for this work looks nothing like that list, be suspicious of it.
None of that means skip it. The work overlaps almost entirely with things that help you in ordinary search and with things that help a human decide to call you. It means don't pay a premium for a new acronym on top of work you should already be getting.
How an AI visibility engagement runs.
Measure first, because otherwise there is no way to tell whether any of it worked.
Baseline
I run your buyers' real questions through ChatGPT, Perplexity, Claude and Google AI Overviews and record who gets named and which sources get cited. That transcript is the before picture.
Find the gap
Usually one of four things: you are unreadable, your facts disagree with each other, nobody else vouches for you, or the question you want to own has no page that answers it.
Build
Structured data, entity cleanup across every place you appear, question-shaped content, and getting listed where these systems already look.
Re-sample
Same prompts, same assistants, on a schedule. You see the transcripts, including the ones where you still do not appear.
Baseline
I run your buyers' real questions through ChatGPT, Perplexity, Claude and Google AI Overviews and record who gets named and which sources get cited. That transcript is the before picture.
Find the gap
Usually one of four things: you are unreadable, your facts disagree with each other, nobody else vouches for you, or the question you want to own has no page that answers it.
Build
Structured data, entity cleanup across every place you appear, question-shaped content, and getting listed where these systems already look.
Re-sample
Same prompts, same assistants, on a schedule. You see the transcripts, including the ones where you still do not appear.
Questions I get about AI search.
What is answer engine optimization?
What is generative engine optimization, and how is it different from AEO?
How do you measure something with no rankings?
Should I block AI crawlers?
What is HEO?
Is this replacing SEO?
Is it worth it for a small local business yet?
Want to know what ChatGPT says about your industry?
I'll run your buyers' real questions through the major assistants and send you what comes back, including who gets recommended instead of you. That part costs nothing.
An honest conversation about your goals, and a straight answer either way.