Generative engine optimization: a small business guide

Say a customer mentions, almost in passing: "I asked ChatGPT who to hire, and it named two other shops."

Picture that evening. You sold all week. You did not know you had to audition for a chatbot. Now you are wondering whether your business is even in the answer, and whether anything you could actually do would change it.

Something can. It is called generative engine optimization, or GEO: the work of making your business easy for AI answer engines to find, understand, name and cite. Some of it is one evening of plain-text fixes. Some of it is a habit you check once a month. None of it is a guarantee, because search engines and answer engines decide what they show, and no one who tells you otherwise can prove it.

This guide covers what GEO actually is, the five fixes worth one evening, and a monthly check you can run with nothing but a list of real buyer questions, plus the limits that the pages ranking for this term mostly skip.

What generative engine optimization actually is

When someone asks ChatGPT, Google's AI Overviews, Gemini or Perplexity a question like "who should I hire to refinish a deck in my city?", the engine does not hand back a list of ten links. It writes one answer, assembled from a handful of sources it chose.

Search engine optimization was a competition to rank at the top of that list. Generative engine optimization is a competition to be part of the answer. Semrush's practical guide puts it cleanly: in SEO you compete to rank at the top, in GEO you compete to be part of the output.

That changes what "winning" looks like. In search, ten businesses get a slot. In an AI answer, maybe three get named, and the rest do not exist that day. If you are named, you are in the consideration set. If you are not, a buyer may never see you at all.

Two things follow from that, and everything else in this guide comes from them:

1. Your website's job changes. A results page forgives a messy page; the visitor clicks the next link. An answer engine has to lift your facts cleanly or it uses someone else's. Clear, quotable pages win.

2. What other people say about you matters more than what you say about you. Every business calls itself reliable on its own site. An engine deciding who to recommend leans on independent signals: press, directories, reviews, mentions on sites you do not control.

Neither of those is a trick. Both are things owners used to do before they were told they were doing SEO.

Why the answer names other shops (and it is not the size of their budget)

In September 2025, researchers at the University of Toronto published a large-scale controlled comparison of AI search against Google. The paper, Generative Engine Optimization: How to Dominate AI Search (arXiv 2509.08919), ran controlled experiments across multiple verticals, languages and query paraphrases. Its central finding: AI search shows a systematic and overwhelming bias toward earned media, meaning third-party authoritative sources, over brand-owned and social content. Google's results mix sources far more evenly.

Read that again from a small business owner's chair. Google balanced things out. The AI engines tilt toward the world's press, roundups and directories, and away from your own pages. The researchers also flag a big-brand bias that niche players have to work to overcome.

This is the finding the thin agency posts ranking for this term do not use, and it is the one that should set your priorities. Writing more pages about yourself on your own site is the weakest move available. Getting named on sources the engines read is the strongest.

Google itself now publishes official guidance on this. Its Search Central guide, Optimizing your website for generative AI features on Google Search (May 2026), says its AI features are rooted in the same core ranking and quality systems as regular Search. So the fundamentals still apply: a technically sound, crawlable site with helpful content is the entry ticket. Google also says what you can stop doing: no special AI rewriting, no content chunking, no special schema or Markdown versions of pages, and llms.txt files get no special treatment.

So the playbook is shorter than the industry sells it as. Make your site plainly readable and quotable, then spend your remaining effort on mentions you do not control. The five fixes below do exactly that.

Where AI answers get their facts about a local business

Before the fixes, know what you are feeding. An AI answer about a local business assembles its facts from a handful of places, and each one is yours to check.

Your own website. Hours, services, service area, prices and the questions customers ask, written as text the page serves. If a fact lives only in a picture or on a phone call, the answer cannot lift it.

Your Google Business Profile and your other listings. Name, address, phone, hours, categories, services and photos. For Google's AI features this is the most local record there is.

Review sites and directories. Reviews supply the description other people write about you, which the answer trusts more than your own page.

Mentions you do not control. Local press, chamber-of-commerce pages, industry directories and community sites. The research above says AI search leans hardest on exactly these.

The rule that ties them together: keep the name, address, phone, hours and one-line description identical on every one of them. Answer engines read across sources, and inconsistency reads as uncertainty.

The five fixes, one evening

None of these needs a tool, a plugin or a consultant. A laptop and two hours get most of it done. The checklist card above is this section, ready to save.

1. Put hours, prices and service area in plain text

Answer engines assemble an understanding of your business from whatever they can lift. If your hours live in an image of a sign, your prices only appear on a phone call, and your service area is implied by a photo, the engine has nothing to quote and nothing to verify.

So write the boring facts as text, on a page any visitor could also use: hours, the cities or neighbourhoods you serve, what a typical job costs or starts at, what you do and what you deliberately do not do. Plain sentences beat clever taglines here. "We refinish decks in Hamilton, Burlington and Oakville" is machine-readable and human-readable, which is the point.

Google's own guide says the same best practices that work for Search work for its AI features. Text it can read and index is where that starts.

2. Answer the question in the first sentence

Engines lift passages, not whole pages. A passage is liftable when a heading is phrased the way a customer asks it, and the first sentence under it is a complete answer on its own.

Take the ten questions customers actually ask you, before they hire you. Do you warranty your work? How long does a job take? What is included in the price? Give each a clear heading and a first sentence that answers it completely, then the detail underneath. An FAQ page is the easiest version of this, and it serves human visitors, traditional search and answer engines identically.

The test is simple: read your first sentence with the heading cut off. If it still answers the question, it can be quoted.

3. Earn mentions you do not control

This is the fix the ranking pages miss, and the research says it is the big one. AI answers lean on third-party sources, so one real mention on a site the engines read can outweigh another page on your own.

You do not need press agents. The earned media available to a small business is unglamorous and real: your chamber of commerce and industry directories, a supplier's or partner's client page, a local journalist looking for a quote on a story in your field, a podcast in your trade, a well-earned base of customer reviews with specific detail in them. Pick two you could realistically earn this month and go after them.

One warning, from Google's guide: seeking inauthentic mentions to influence what is said about you is not likely to help, because its AI features rely on the same spam-defence systems as core Search. Earn the mention. Do not manufacture it.

4. Keep your facts identical everywhere

Your website says one phone number, your Facebook page another, a directory an old address. To a human this is untidy. To an answer engine assembling a picture of your business from many sources, inconsistency reads as uncertainty, and uncertain businesses do not get recommended.

Pick the correct version of your business name, address, phone number, hours and one-sentence description of what you do. Then correct every listing you can find, starting with the places customers actually see: Google Business Profile, Facebook, Instagram, and the main directories in your industry.

This is also where the monthly check below earns its keep. When something changes, hours in winter, a new service area, change it everywhere in the same sitting.

5. Be specific instead of commodity

The engine's answer names a shortlist. Commodity businesses look interchangeable, so the shortlist gets filled by whoever is most described. "We do landscaping" is interchangeable. "We rebuild heritage front porches in Toronto's older neighbourhoods, cedar and Victorian detailing" is a thing an engine can match to a question about heritage porches.

Write pages about the specific problems you solve, the specific materials and methods you use, the specific customers you are wrong for, and the decisions a buyer faces. Depth on one niche beats a thin page about everything, and it is exactly how a niche player works against the big-brand bias the research describes.

The monthly check, no tools

You cannot improve what you never look at, and you do not need software to look. Here is the method this guide proposes. Once a month, roughly fifteen minutes, do this.

Write the ten questions a real buyer would ask an assistant before hiring you. Two are discovery ("who should I hire to..."), a few are comparison ("is X or Y better for..."), a few are trust ("what should I know before hiring a..."), and one is about you by name ("what can you tell me about [your business]?").

Ask them in ChatGPT and in Google, or wherever your customers actually ask. For each question, log three things in three columns:

Mentioned: is your business named at all?

Recommended: are you named as an option or the answer?

Cited: is your website used as a source, with a link?

That log is your scoreboard. Keep the same ten questions every month so the comparison means something, and change a question only when your service changes.

Watch the log for direction, not for a promise. Names appearing and citations moving month over month tell you the work is landing. A single bad week tells you nothing, because engines update their models and patterns shift with any model change.

What nobody can promise

Here is the paragraph the pages ranking for this term mostly skip.

No one can guarantee you a place in an AI answer. Not an agency, not a plugin, not us. Search engines and answer engines decide what they show, and their systems change without notice. The same Toronto paper that mapped AI search's preferences also found the engines differ sharply from each other in domain diversity, freshness and sensitivity to phrasing, which means there is no single knob to turn, and findings can shift with the next model update. Google's guidance says its AI features use the same core ranking and quality systems, so a site that does not rank in regular search should not expect AI answers to rescue it.

So treat anyone who quotes you a guaranteed placement or an exact timeline accordingly. Our own help page on search and answer engines (https://nouna.io/help/seo-geo-and-evidence) states the rule we work by: the checks and scores tell you where you stand and what to fix, not what will happen.

What makes these five fixes a safe bet is that they pay even if every AI engine vanished tomorrow. Plain facts, clear answers, real mentions, consistent details and a specific offer are simply good marketing. GEO is the habit of doing the marketing you should have been doing anyway, in a format a machine can quote.

If you would rather not run it yourself

The evening fixes are the easy half. The work is doing the checks, applying the fixes, keeping facts current and earning mentions every month after, which is usually where owners run out of hours.

Nouna does this as standing work: a search and AI-answers specialist runs fixes on your site every month, 9 on Starter, 17 on Growth, 30 on Department. The plan pages (https://nouna.io/pricing) list what each size makes, from $149 USD a month. It never promises what engines decide, and site changes beyond small fixes wait for your yes.

If you want to see the work on your own business first, start with the free website read (https://nouna.io/start), no card: your website analysis, your six strategy documents, your search audit and 20 free credits to spend on any channel.

And if budget is the question behind the question, start from what one customer is worth, not from a percentage. That rule, and how to test it in two weeks, is in our piece on what to spend on marketing (https://nouna.io/blog/how-much-to-spend-on-marketing). And if you are weighing an agency against a department, our guide to what an AI marketing agency costs (https://nouna.io/blog/ai-marketing-agency-what-it-is-what-it-costs) has the 2026 prices.