Patrick J. Buckhoff, CCIM, runs RE/MAX Commercial Advisors Group in Houston. In 2026 we restructured the brokerage's site content for AI answer engines. Afterward, Patrick told us this:
"Recently I've gotten three calls inbound that people found us online. Two from chat, one from Claude."
Two from chat. One from Claude. Three inbound calls where an AI assistant, somewhere along the way, put the brokerage in front of the buyer. No ad. No referral. No cold email.
That's generative engine optimization. It's the work of getting your firm named, quoted, and linked when a buyer asks an AI assistant for help. And for founder-led B2B service firms, it's quickly becoming the difference between being on the shortlist and never knowing a shortlist existed.
This is the whole playbook. What GEO is, how it differs from the SEO you already know, the framework we use to make firms quotable, how to test where you stand today, and what to stop doing.
What is generative engine optimization (GEO)?
Generative engine optimization (GEO) is the practice of shaping your content and your reputation so generative AI assistants like ChatGPT, Claude, Perplexity, Gemini, and Google's AI Mode mention your firm, describe it accurately, and cite your pages when they answer a buyer's question. Traditional SEO competes for a ranking on a results page. Answer engine optimization (AEO) competes to be the direct answer in featured snippets, voice assistants, and Google's AI Overviews. GEO competes for a place inside the conversation a buyer has with an AI assistant.
The term comes from a 2024 research paper out of Princeton and IIT Delhi, presented at KDD 2024, which found that specific content changes can boost visibility by up to 40% in generative engine responses. More on what those changes were in a minute. They're not what most SEO vendors are selling you.
Why this matters for a B2B service firm right now
Your buyers have always done their homework before calling you. What changed is where the homework happens.
Buyers make up their minds early. The 6sense 2024 Buyer Experience Report found B2B buyers are nearly 70% through their purchasing process before they engage a seller, and 81% already have a preferred vendor by the time they make first contact. By the time they call you, the decision is mostly made.
More of that early homework now happens in a chat window. Responsive surveyed 350 B2B buyers in October 2025 and found generative AI has overtaken traditional search for a quarter of B2B buyers, with nearly two-thirds using it as much as or more than search when researching vendors.
And when people do use Google, the AI summary is eating the click. Pew Research Center found Google users who saw an AI summary clicked a traditional result in 8% of visits, compared with 15% when no summary appeared. They clicked a link inside the summary itself in just 1% of visits.
Gartner saw this coming back in February 2024, when it predicted traditional search engine volume would drop 25% by 2026 because of AI chatbots and virtual agents.
Put those together and you get a new buying reality for a consulting firm, an M&A advisory, an executive search firm, an accounting practice. A CFO with a problem asks an AI who handles it. The AI gives a short answer with three or four names. The CFO researches those names. One of them gets the call.
If you're not in the answer, you're not in the conversation. And unlike a lost Google ranking, you won't even see the loss. Nothing shows up in your analytics. The meeting just never happens.
How AI answers differ from Google rankings
If you've done any SEO, your instincts will mislead you here in a few specific ways.
There is no position one. Google gives you a ranked list. An AI assistant gives you a paragraph. You're either named in it or you're not, and if you are, the words around your name matter as much as the mention.
The answer changes every time. SparkToro and Gumshoe.ai ran 12 identical prompts through ChatGPT, Claude, and Google's AI, again and again. Across about 3,000 runs, the odds of getting the same list of brands twice were under 1 in 100, and the same list in the same order closer to 1 in 1,000. Some brands still showed up in 60% to 90% of answers for a given question. That's the real goal: showing up so often that the randomness stops mattering.
The AI is summarizing the web, including the parts you don't own. When an assistant answers "who are the best sell-side advisors for HVAC companies in Texas," it pulls from your site, yes. It also pulls from directories, association pages, podcast show notes, LinkedIn, press coverage, and reviews. What other credible sources say about you counts heavily.
Accuracy is now your problem. Google linked to your page and let you speak for yourself. An AI paraphrases you. If your site is vague, it fills the gaps with guesses. Picture a boutique sell-side advisory described as a "full-service investment bank" because its homepage never said otherwise.
The click is optional. A buyer can get your name, your specialty, and your best result without ever visiting your site. So the page has to do its job even when nobody lands on it.
The good news: these systems still build many of their answers from pages they retrieve from the web. Good SEO hygiene, a fast crawlable site, clear pages, real links, is still the foundation. GEO sits on top of it. We break down exactly how the three disciplines overlap in GEO vs SEO vs AEO.
Two ways an AI learns about your firm
It helps to know where the answer comes from, because it changes what you fix first.
Route one is memory. Every large language model was trained on a huge snapshot of the web. If your firm was described clearly and often across credible sites before that snapshot, the model may know you without looking anything up. You can't edit that memory directly. You influence the next version of it by being described the same way, in many places, over time.
Route two is a live search. ChatGPT, Claude, Perplexity, and Google's AI features can all go out and pull current pages from the web to build an answer, then cite them. This is the route you can move fast. Clear claims, clean pages, and proof on your own site show up here within weeks of being published and crawled.
So the work runs on two clocks. The live-search clock rewards what you publish this month. The memory clock rewards what the whole internet has said about you for years. A good GEO program feeds both: tight, specific pages on your own site for the fast route, and a steady drumbeat of mentions, appearances, and published thinking for the slow one.
What GEO looks like for different kinds of firms
The playbook is the same. The questions change.
M&A and financial advisory. Owners ask, "How do I sell my company?" and "Who are the best sell-side advisors for [industry] companies in [state]?" Your sector focus, deal size range, and named transactions are the claims that matter.
Executive search. Boards and CEOs ask who places leaders in their function and industry. Specialty, seniority level, and time-to-placement results carry the answer.
Accounting and advisory practices. Owners ask about specific situations: "Who handles R&D tax credits for manufacturers near me?" Service pages built around those exact situations win.
Commercial real estate. Owners ask for a broker who knows their property type and submarket. RE/MAX Commercial Advisors Group publishes deep, specific content on Houston submarkets, loans coming due, and 1031 exchanges.
Consulting and coaching. Buyers ask how to fix a problem before they ask who fixes it. The firm that published the clearest answer to the problem tends to get named as the one to call.
Whatever you do, start by writing down the ten questions buyers ask before they know your name. Those questions are your GEO plan.
The CITE Framework: how to become the firm AI quotes
After restructuring client sites for AI answer engines, including RE/MAX Commercial Advisors Group's, we boiled the work down to four letters. CITE.
- Claim. Say the specific, checkable thing.
- Instrument. Make the page easy for machines to read.
- Testimony. Show proof from named clients and third parties.
- Evidence. Cite named sources, with links and dates.
Each letter answers a question the AI is quietly asking: What exactly does this firm do? Can I read this page cleanly? Does anyone else vouch for it? Can I trust these facts?
C is for Claim: say the specific thing
Write your website to be repeated. That's what AI rewards.
Here's a sentence from a real-feeling homepage: "We deliver tailored strategic solutions that help growth-minded organizations unlock their full potential."
An AI can't do anything with that. There's no who, no what, no where. It can't quote it because there's nothing in it worth quoting.
Now try this: "We run sell-side M&A for founder-owned HVAC and mechanical services companies with $10 million to $50 million in revenue in Texas and Oklahoma."
Every word is a fact. Service. Client type. Industry. Size. Geography. When a buyer asks for exactly that, the model has a sentence it can lift almost whole.
Write your claims so they survive being copied out of context. Who you serve. What you do for them. Where. What happened when you did it, with numbers. Put that sentence on your homepage, your about page, your LinkedIn company page, and every directory profile you control, worded the same way each time. Consistency is how a model learns to trust a fact.
The Princeton research backs the instinct. The content changes that lifted visibility most were adding quotations, adding statistics, and citing sources. Keyword stuffing performed worse than doing nothing at all on the study's main visibility measure. Specific, sourced content wins.
I is for Instrument: make the page machine-readable
The second letter is plumbing. Unsexy, and it matters.
Structure every important page around the questions buyers ask. Question-style headings. A direct two-to-four sentence answer right under each one. Then the depth. That pattern is easy for a model to extract and easy for a skimming human to love.
Add structured data (schema markup) that states your facts in a format machines read cleanly: Organization, ProfessionalService, Person for your partners, Service, Article, FAQPage. One honest note here, because a lot of people oversell this. Google says plainly that no special schema is required to appear in AI Overviews or AI Mode, and that your structured data should match the visible text on the page. Schema makes a specific page unambiguous. The specifics still have to be there first. We walk through every type a B2B firm needs in Structured Data for Answer Engines.
Then check the basics most firms never look at. Is your real content in the HTML, or loaded by a script that some crawlers skip? Does your robots.txt let AI crawlers like GPTBot, ClaudeBot, and PerplexityBot in? Are your case studies trapped in a PDF or a slide image? A page a model can't read is a page that doesn't exist.
T is for Testimony: let others vouch for you
You can say anything about yourself. Models know that. So they weigh what other credible sources say.
Testimony starts at home with named client results. A case study that says "a client in financial services" is weak. A case study that says Oaklyn Consulting grew gross profit 93% year over year is something a model can repeat with a name attached. Named results are the single most quotable asset a service firm owns.
Then go outside your site. Guest appearances on podcasts your buyers listen to. Speaking pages on association websites. Directory listings with your specialty spelled out. Quotes in trade publications. Reviews on your Google Business Profile. Every credible page that describes you the same way you describe yourself is another vote.
This is the old reputation game, with a new referee. The 2025 Edelman-LinkedIn B2B Thought Leadership Impact Report found 75% of decision-makers said thought leadership pushed them to research a firm they hadn't considered. Now the research step often runs through an AI, and the AI reads the same thought leadership.
E is for Evidence: show your sources
The last letter is provenance. Where did this fact come from, and can anyone check it?
When you cite a number, name the source and link it. Date your articles. Put a real author byline on them with real credentials. Update your key pages and show the update date. Publish original data when you have it: a survey of your clients, a benchmark from your own deals, a pattern you've seen across fifty searches.
Every stat in this article has a named source and a link for exactly this reason. Showing your work is how a page earns trust, from a buyer and from a model.
The full anatomy of what gets quoted, and what gets skipped, lives in The Content Patterns LLMs Quote.
How to test your brand in ChatGPT, Claude, and Perplexity
Before you fix anything, find out where you stand. It takes about three minutes.
Open ChatGPT, Claude, and Perplexity in a private or logged-out session, with memory off, so your own history doesn't tilt the results. Then ask:
- The identity question. "What does [your firm] in [your city] do, and who do they serve?" This tests whether the AI knows you exist and gets your facts right.
- The category question. "Who are the best [your service] firms for [your client type] in [your region]?" This tests whether you make the shortlist.
- The problem question. Ask the thing a buyer asks before they know your name. "How do I sell my HVAC company?" "How do I find a CFO for a PE-backed manufacturer?" This tests whether your expertise shows up at the moment of need.
- The proof question. "What results has [your firm] gotten for clients?" This tests whether your case studies are reachable and quotable.
- The source question. "What sources did you use for that?" This shows you exactly which pages are doing the work, and which ones aren't.
Run each prompt at least three times in each engine. Remember the SparkToro finding: one answer is a coin flip. A pattern across runs is information.
Write down what you see. Named or not named. Accurate or wrong. Your pages cited or someone else's. That's your baseline. The full version, with every prompt and what a good answer looks like, is in The Prompt-Test Your Brand in Three Minutes Checklist.
What SEO moves still work, and which ones to retire
Good news first. Most of what made a firm findable on Google still helps it get named by AI. The overlap is bigger than the hype suggests.
Keep doing these.
- A fast, crawlable site with clean page titles, descriptions, and internal links.
- Topic clusters: one deep pillar page plus focused supporting articles that link to each other. Depth on one subject builds authority on it.
- Earning links and mentions from credible sites in your industry.
- A complete, consistent Google Business Profile.
- Publishing on a schedule. Freshness and volume of genuinely useful content still compound.
Retire these.
- Keyword stuffing. The Princeton team found it underperformed doing nothing on their main measure.
- Thin "ultimate guide" pages that say what every other page says. A model has no reason to quote the 400th copy of the same advice.
- Gating your best thinking behind a form. A PDF behind an email wall is invisible to every AI assistant your buyer uses.
- Writing for a keyword at the expense of a human. "Best consulting firm Chicago consulting services Chicago" reads like spam to people and gives a model nothing true to repeat.
- Obsessing over rank position as the only scoreboard. You now need a second scoreboard for how often AI names you and how accurately.
Add these.
- A one-sentence claim about who you serve, repeated consistently everywhere you're listed.
- Question-led page structure with direct answers.
- Named client results with percentages.
- Named, linked sources for every stat you use.
- A monthly prompt test, logged in a simple spreadsheet.
The twelve moves that compound fastest are laid out step by step in The AEO Checklist for B2B Service Firms.
A worked example: one firm, ninety days of CITE
Make it concrete. This is a hypothetical, but every move in it is real.
Say you run a 12-person executive search firm in Atlanta that places CFOs and controllers at private equity-backed manufacturers in the Southeast. You're good. Your clients love you. Almost all your work comes from referrals.
You run the prompt test. ChatGPT describes you as "a recruiting agency offering staffing solutions across industries." Claude doesn't know you. Perplexity names three national firms and a local staffing company when asked about CFO search for PE-backed manufacturers in the Southeast. You're nowhere.
Weeks 1 to 3: Claim. You rewrite the homepage opening to one sentence: "We recruit CFOs and controllers for private equity-backed manufacturers across the Southeast." Same sentence goes on your about page, LinkedIn company page, and every directory profile. Each partner's bio names the functions and sectors they place, with years in the seat.
Weeks 4 to 6: Instrument. Your service pages get question headings: "How long does a CFO search take for a PE-backed manufacturer?" with a direct answer under each. Your developer adds Organization, ProfessionalService, and Person schema that matches the page text. Someone finally checks robots.txt and finds AI crawlers were blocked by a security plugin. Fixed.
Weeks 7 to 9: Testimony. You publish two case studies with client permission, each with a named company and percentages: time to placement, retention at 18 months. One partner goes on a private equity operations podcast. You update your profile on the association site your clients belong to, using the same claim sentence.
Weeks 10 to 12: Evidence. You publish a short report built from your own data: what PE-backed manufacturers in the Southeast actually look for in a first CFO, drawn from your last 40 searches. Dated, bylined, sourced. It's the only page on the internet with that information.
Run the prompt test again in month four. In this example, ChatGPT now describes you correctly. Perplexity cites your report when asked what PE firms want in a CFO. Claude names you in about half its runs for the category question. Every piece of that was built on purpose.
That last piece, the report, is where most firms leave the biggest win on the table. You're sitting on proprietary insight from years of real work. Publishing it is the most citable thing you can do.
How to measure GEO without fooling yourself
The first time a partner pastes a ChatGPT answer that mentions your firm into the group chat, everyone cheers. Enjoy it. Then ignore it, because one answer proves nothing.
Measure what holds up across many runs:
- Mention rate. Of all the times you run your fixed prompt set across engines, what share name you?
- Share of voice. How often do you appear compared with three to five named competitors?
- Citation rate. How often does the answer link to your own pages, and which pages?
- Accuracy. Are the facts the AI states about you right?
- AI-referred pipeline. Sessions from chatgpt.com, perplexity.ai, claude.ai, and similar sources in your analytics, plus the answer to "how did you hear about us" on every intake form and discovery call.
That last one matters most. Patrick at RE/MAX Commercial Advisors Group could name the source of each of those three calls: two "from chat," one "from Claude." Traffic is a weak signal when AI summaries keep buyers from clicking. A booked meeting is a strong one.
Some early data suggests AI-referred visitors are worth chasing. Ahrefs reported that on its own site, AI search sent 0.5% of visitors but drove 12.1% of signups, converting at 23 times the rate of traditional organic visitors. That's one company's data, so hold it loosely. The direction is still worth noticing: buyers who arrive from an AI answer have already been pre-sold.
The full scorecard, with the arithmetic, is in How to Measure AI Visibility in 2026.
What actually stops firms
Three things get in the way. None of them is technical.
The founder is the only one who can say the specific thing. The sharpest claims, the best stories, the real numbers all live in one very busy head. So the website stays vague, because vague is what a marketing coordinator can write without bothering you. Fixing this means getting your thinking on the record once and reusing it everywhere.
Proof sits behind NDAs and modesty. Most firms have more client results than they think, and publish almost none. Ask for permission. Use percentages when absolute numbers are sensitive. Anonymize where you must. A named result beats an anonymous one, and an anonymous one beats nothing.
It feels like a fad. I get it. Every year brings a new acronym. Here's the honest test: everything in the CITE Framework also makes your firm more persuasive to a human reading your site. Specific claims, clean pages, real proof, sourced facts. If AI search vanished tomorrow, you'd still have a better website. There's no scenario where this work is wasted.
The GEO and AEO cluster: every guide in this playbook
This pillar is the overview. Each piece below goes deep on one part of the work.
- GEO vs SEO vs AEO: What Each One Means and Why They're Converging. The plain-English definitions, side by side, and why you should run them as one discipline.
- How to Make Your Website Citable by ChatGPT, Claude, and Perplexity. Specificity, structured data, and provenance, with before-and-after rewrites.
- The AEO Checklist for B2B Service Firms (12 Moves That Actually Compound). The scannable list, organized by the CITE Framework.
- Structured Data for Answer Engines: The Schema Types B2B Firms Need. Which schema types matter, a full example, and how to get Claude to draft yours.
- The Content Patterns LLMs Quote (and the Ones They Skip). The anatomy of a sentence an AI wants to repeat.
- How to Measure AI Visibility in 2026. A five-metric scorecard that holds up to scrutiny.
- The Prompt-Test Your Brand in Three Minutes Checklist. The exact prompts to run today, and what each one tells you.
Where to start this week
You don't need a six-month project to begin. You need an hour and some honesty.
Run the prompt test. Write your one-sentence claim and put it on your homepage. Pick your best client result and get permission to name it. Check your robots.txt. That's the first week.
Then keep going, every month, on purpose. Generative engine optimization rewards the firm that publishes specific, proof-backed thinking on a schedule. The firm owners ask an AI about in 2027 is being built right now, one clear page at a time.
RE/MAX Commercial Advisors Group put years of specific, useful market thinking in public, and then we made sure the machines could read it. Three calls later, Patrick had his answer.
Frequently asked questions
What is generative engine optimization (GEO)? Generative engine optimization is the work of getting your firm named, quoted, and linked inside answers from AI assistants like ChatGPT, Claude, Perplexity, and Google's AI Overviews. SEO earns a spot on a results page. GEO earns a spot inside the answer itself.
Is GEO replacing SEO? No. AI assistants build many of their answers from pages they retrieve from the web, so a crawlable, well-structured, well-linked site is still the foundation. GEO adds a second scoreboard on top of SEO: whether the answer names you and gets your facts right.
How do I know if ChatGPT recommends my firm? Ask it the questions your buyers ask, in a private session, and run each prompt several times across ChatGPT, Claude, and Perplexity. AI answers vary from run to run, so one screenshot proves nothing. Track how often you appear across many runs.
What is the CITE Framework? CITE is Rockstarr & Moon's framework for AI-discoverable content: Claim (say the specific, checkable thing), Instrument (make the page machine-readable), Testimony (show real proof from named clients and third parties), and Evidence (cite named sources with links and dates). Each letter makes your content easier for an AI to quote with confidence.
How long does GEO take to work? The fixes to your own site can be done in weeks. Showing up consistently in AI answers usually takes months, because the models lean on what the rest of the web says about you too. Firms that publish specific, proof-backed content every month compound fastest.
Want to know what ChatGPT, Claude, and Perplexity say about your firm today, and what to fix first? Let's look together.