GEO

How to Choose a GEO Agency: 10 Questions, Red Flags, and What Good Looks Like

Ned MehicNed MehicFounder, OrkkidAugust 14, 2026Updated: October 7, 202612 min
How to Choose a GEO Agency: 10 Questions, Red Flags, and What Good Looks Like

How to choose a generative engine optimization (GEO) agency without buying repackaged SEO: 5 signs of relabeled services, 10 questions that expose real capability, red flags that should end a sales call, and realistic results month by month.

Choosing a GEO agency comes down to one test: can they show you, with repeatable measurement, how often AI engines name your business today and how they will change that? A real generative engine optimization agency tracks citations against a fixed panel of customer prompts, fixes AI crawler access and structured data, restructures content so it can be quoted, and reports honestly, including the prompts where you do not appear. An agency that cannot explain those things is selling SEO with a new label.

That is the real risk right now. Many agencies suddenly sell GEO, AI SEO, or AEO. The same providers who pitched "SEO packages" two years ago now have an "AI search optimization" page, and from the outside you cannot tell who understands how generative engines work and who renamed a service and raised the price. The danger is not paying for GEO and getting nothing. It is paying for GEO and getting 2019 SEO.

This guide is the filter. We wrote it as a GEO agency ourselves, which means two things: we know what the work involves, and yes, we have an interest. So the whole post is built on questions and evidence you can verify independently, whether you shortlist us or never talk to us. If you want the background first, our guide to what generative engine optimization is explains the discipline, and GEO vs SEO covers how it differs from what your current agency may already do.

One honest way to test any agency, including us: request a free AI visibility audit and compare what comes back against the standards below.


TL;DR

  • In our October 2026 test of 120 AI answers across eight cities, two AI engines asked the same question agreed on only 17% of their recommended businesses on average, and the most-cited source differed by industry (movebuddha.com for movers, weddingwire.com for wedding photographers, energysage.com for solar installers), so ask every agency to show multi-engine tracking and a plan for the review sites AI reads in your niche.
  • The biggest risk when hiring a GEO agency is paying a premium for old SEO with new labels. The tell is deliverables: if nothing in the proposal mentions citation tracking, prompt panels, entity work, or AI crawler access, it is repackaged SEO.
  • Ask ten specific questions before signing: how they track citations, what their prompt panel looks like, how they handle llms.txt and schema, which engines they prioritize, and what their realistic timeline is.
  • Walk away from guaranteed placement in ChatGPT, cherry-picked screenshots as proof, and any engagement that starts without a baseline measurement.
  • Fair GEO reporting shows citation share against a fixed prompt panel, month over month, across multiple engines, including the prompts where you do not appear.
  • Realistic results: baseline and technical foundation in month one, first citation movement by months two to three for established sites, meaningful citation share by months four to six. Faster claims deserve skepticism.
  • Specialist agencies that focus on one industry often have an edge, because they already know the prompts, sources, and review platforms that matter in that market. Test that claim like any other.

Why choosing a GEO agency is harder than choosing an SEO agency

SEO has two decades of established norms. You can check an SEO agency's own rankings, audit their backlinks, and benchmark their pricing against a mature market. GEO has none of that maturity. The discipline is young, the tooling is new, there is no long track record of case studies, and the engines themselves change every few months. That vacuum is exactly where overclaiming thrives.

It also means the evaluation burden shifts to you. You cannot rely on the market to filter out the pretenders, because it has not yet. What you can do is understand what real GEO work consists of, then test every provider against it. The work itself is not mysterious. We lay our method out publicly on our generative engine optimization page, and a legitimate agency should be able to explain its version without hiding behind vague, proprietary-sounding language.

The 5 signs a "GEO service" is just old SEO relabeled

These are the patterns we see most when reviewing proposals that prospective clients bring us for a second opinion.

1. The deliverables list is a keyword deliverables list

Read the proposal's actual line items. If they are "keyword research, on-page optimization, monthly blog posts, backlinks, ranking report", that is an SEO retainer, whatever the cover page says. Real GEO deliverables include a citation baseline, prompt panel design, AI crawler and llms.txt configuration, entity and schema markup work, content restructuring for citability, and citation share reporting. Some SEO line items belong in a GEO scope too, because the disciplines share a foundation. But if none of the AI-specific items appear, nothing AI-specific is being done.

2. They cannot explain how they measure AI visibility

Ask "how will I know it is working?" and listen. If the answer is rankings and traffic, they are measuring SEO. AI answers often produce no click at all, so an agency measuring only clicks is blind to the thing you are paying for. The right answer involves citation share: running a fixed panel of real customer prompts across engines each month and reporting how often you are named. Our guide on how to track ChatGPT citations shows what that looks like in practice.

3. Every engine is treated as one thing called "AI"

ChatGPT retrieves through Bing. Google AI Overviews and Gemini read Google's index. Perplexity runs its own crawler and shows sources prominently. An agency that talks about "ranking in AI" as a single activity has not done the work at engine level. You do not need a lecture on architecture, but they should be able to tell you which engines matter most for your market and why the tactics differ.

4. No interest in your entity or your wider web presence

GEO is heavily about whether engines can verify your business as a consistent, trustworthy entity across your site, your Google Business Profile, directories, reviews, and brand mentions elsewhere. A proposal that never leaves your website is doing at most half the job.

5. The price went up but the team did not change

A fair question for any provider who added GEO to the menu recently: what did you change? New tooling, new process, staff who specialize in it, published thinking on the topic? "We have always done this" is an honest answer only if their old deliverables genuinely covered it, which you can now check against sign number one.

→ Not sure where you stand? Our free GEO audit tests the questions your customers ask and shows you the answer. Run my free audit.

The 10 questions to ask a GEO agency

Take these into every sales call. The point is not to catch anyone out with trivia. Each question has answers that separate practitioners from packagers.

  1. "Show me a citation tracking report for an existing client, anonymized." The single best question. A real practitioner has these and will show you the format in two minutes. A packager will describe a report that does not exist yet.
  2. "How do you build the prompt panel for a business like mine?" Good answer: from your real customer questions, inquiries, and service lines, fixed at baseline so results are comparable month to month. Bad answer: vague talk of "monitoring AI mentions".
  3. "What is your baseline process before work starts?" Non-negotiable. If they do not measure where you stand across engines before starting, every future claim of improvement is unverifiable.
  4. "Which AI crawlers will you check and configure, and will you implement llms.txt?" Listen for specifics: GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Bing indexing. This is basic technical GEO hygiene. You can check your own site first with our free AI crawler checker.
  5. "What schema and entity work is included?" You want depth here: Organization, LocalBusiness, Service, FAQPage, Article, plus consistency work across your profiles and listings, not just "we add schema".
  6. "Which engines will you prioritize for us, and why?" Tests whether they think at engine level and whether they have thought about your market at all.
  7. "How does your GEO work interact with our existing SEO?" The right answer is that they share a foundation and compound each other. An agency that says SEO is dead is selling a story, not a strategy.
  8. "What does a realistic timeline look like for a site like ours?" Compare their answer with the month-by-month expectations later in this post. Wild promises here predict wild reporting later.
  9. "What happens to the work if we leave?" Schema, content, llms.txt, and profile improvements should live on assets you own. Beware anything that lives in a proprietary platform that leaves with the agency.
  10. "What do you publish about GEO, and about our industry?" Practitioners tend to have public thinking: guides, breakdowns, tools, case studies. It is not mandatory, but an agency selling expertise with zero published evidence of it is asking for a lot of trust.

Two tests you can apply, backed by our own data

In October 2026 we asked ChatGPT, Perplexity, Gemini, Claude and Google AI Mode to recommend moving companies, wedding photographers and solar installers in eight cities, 120 answers in all. Orkkid sells GEO services, so we have an interest in these findings; the method is at the end of this post and the full results are in our AI recommendation study, so you can check the numbers yourself. Two findings translate directly into questions for any agency, including us.

Test 1: Ask to see tracking across several engines

How much the AIs agree with each other

Average overlap between two engines' shortlists for the same city and industry (100% = identical lists)

Ask two different AIs the same question and their shortlists overlap by 17% on average.

ChatGPTPerplexityGeminiClaudeGoogle AI Mode
ChatGPT100%19%19%26%14%
Perplexity19%100%14%22%8%
Gemini19%14%100%17%12%
Claude26%22%17%100%16%
Google AI Mode14%8%12%15%100%

Source: Orkkid AI Recommendation Study, October 2026 (24 city and industry pairs)

When two engines answered the identical question for the same city, their shortlists overlapped by 17% on average. Of the 558 businesses named, 71% were named by only one of the five engines, and none of the 24 city and industry prompts got the same first pick from all five. An agency that checks ChatGPT alone, or shows you one screenshot from one engine, is reporting on a small slice of what your customers see.

What to ask: "Show me a sample report with my prompts run on at least three engines, with the results per engine." A good answer shows a fixed prompt list, one row per engine, and the misses alongside the hits.

Test 2: Ask which review sites AI reads in your industry

The sites AI reads are different in every industry

Share of answers in each industry that cited the domain (top six per industry)

Each industry has its own gatekeeper: movebuddha.com for movers (47.5% of answers), weddingwire.com for wedding photographers (52.5%) and energysage.com for solar installers (65%).

  • movebuddha.com47.5%

    19 of 40 answers

  • freightwaves.com42.5%

    17 of 40 answers

  • consumeraffairs.com35%

    14 of 40 answers

  • reddit.com30%

    12 of 40 answers

  • extraspace.com30%

    12 of 40 answers

  • forbes.com20%

    8 of 40 answers

Source: Orkkid AI Recommendation Study, October 2026 (120 answers, 40 per industry)

Industry review sites made up 39.6% of everything the AIs cited, more than any other kind of source, and each industry had its own gatekeeper. moveBuddha appeared in 47.5% of moving answers, WeddingWire in 52.5% of wedding photography answers and EnergySage in 65% of solar answers. A plan that only edits your website leaves out the pages the AIs read most often.

What to ask: "Which sites do AI engines cite when customers ask about businesses like mine, and how will you get us listed and reviewed there?" An agency that has done the work will name specific sites for your industry and city without looking them up.

The numbers also show why industry knowledge matters. The three industries behaved very differently:

IndustryDifferent businesses namedNamed by only one engineAverage overlap between two enginesMost-cited site (share of answers)Industry review sites' share of citations
Moving companies16461.6%19.1%movebuddha.com (47.5%)32.5%
Wedding photographers26783.1%5.3%weddingwire.com (52.5%)57.3%
Solar installers12757.5%25.3%energysage.com (65%)28.9%

Wedding photography was the most fragmented market in our test: the engines named 267 different photographers, and two engines shared only 5.3% of their picks on average. An agency that quotes the same playbook and the same timeline for a wedding photographer and a solar installer has probably not measured either market.

Specialist or generalist: does industry focus matter?

You will meet two kinds of GEO agency: generalists who take any business, and specialists who focus on a few industries. Neither is automatically better, but specialists often have a real edge, for practical reasons:

  • They already know the prompts. The questions couples ask about wedding venues are nothing like the questions homeowners ask about solar panels. A specialist starts with a prompt panel that reflects how your customers actually talk.
  • They know which sources the engines trust in your market. Every industry has its own review platforms, directories, and publications that AI engines read. A specialist knows where your mentions need to be.
  • They have comparable baselines. An agency that has measured many businesses in your category can tell you what "good" citation share looks like for your niche.

The trade-off is that a specialist may have less experience outside its niche, and may also work with your competitors. Ask how they handle that. And test the specialist claim like any other: ask for anonymized reports and case studies from your industry, not just a niche-sounding website.

For transparency: Orkkid is a specialist. We focus on moving companies, wedding businesses, and solar installers. That does not make us the right fit for everyone, and the questions above apply to us too.

Free GEO audit

Be the business AI recommends

Run my free audit

Red flags that should end the conversation

Some things you learn about a GEO agency are not "weigh it up" issues. They are walk-aways.

  • Guaranteed placement in ChatGPT or any AI engine. Nobody controls what a generative model outputs. "Position one in ChatGPT" guarantees are the GEO version of the guaranteed-rankings scam SEO spent a decade outgrowing.
  • Screenshots as proof. AI answers vary by phrasing, session, account, and day. A screenshot of one favorable answer proves one favorable answer happened once. Real evidence is repeated measurement against a fixed prompt panel over time.
  • No baseline, straight to retainer. If they will not measure the starting point, they cannot demonstrate progress, and that is usually the intent.
  • Pressure around a closing window. "AI search is moving fast" is true. "So sign this week" does not follow. The opportunity will still be real after you have done due diligence.
  • Secret methodology. There are genuinely proprietary tools in this space, but the method itself is public knowledge. "We cannot tell you what we would do" almost always means "we would not do much".
  • They will not name the work on your invoice. Line items should map to deliverables you can verify: pages restructured, schema deployed, prompts tracked, mentions earned.

What fair GEO reporting looks like

You are entitled to reporting that would survive an audit by a skeptic. A monthly GEO report should show:

  • Citation share against a fixed prompt panel. The same 20 to 50 prompts, run across ChatGPT, Gemini, Perplexity, and Google AI Overviews, with your appearance rate per engine, per month. Including the prompts where you did not appear. Selective reporting is the polite version of the screenshot trick.
  • Accuracy, not just presence. When engines mention you, is what they say correct? Wrong pricing, discontinued services, and stale details in AI answers are findings, not footnotes.
  • Work completed, mapped to the plan. What was shipped this month and which signal it targets.
  • Leading indicators. AI referral traffic, brand mention growth, Bing and Google index health, crawler access status.
  • Honest movement in both directions. Citation share fluctuates as engines update. A report that only ever goes up is a report that is being curated.

If a prospective agency's sample report matches that description, you are probably talking to a practitioner.

Free GEO audit

Get your free GEO audit

Buyers now ask AI for a shortlist before they visit a single website. Find out whether your business is on it, and how to get there first.

Tested across ChatGPT · Perplexity · Google AI Overviews · Gemini

Run my free audit

Realistic results, month by month

Timelines vary with your starting authority, your market, and your competition, but here is the honest shape for an established service business site. Treat large deviations from this, in either direction, as something to question.

PeriodWhat should be happening
Month 1Baseline citation audit, prompt panel fixed, technical foundation shipped: crawler access, llms.txt, schema, indexing in Bing and Google.
Months 2 to 3Core pages restructured for citability, entity consistency fixed across profiles and directories, first content shipped. First citation movement often appears here for established sites.
Months 4 to 6Citation share trending up on the panel, appearing for a growing set of prompts, AI referral traffic measurable. This is where you judge whether the engagement is working.
Months 6 to 12Compounding: topical depth built out, mentions earned across third-party sources, citation share defended as competitors wake up.

New domains sit behind that curve by a few months, because engines need time to learn the entity exists. Broad, highly competitive prompts move slower than specific service questions, where the pool of candidates is small. An agency that sets these expectations unprompted is showing you how it will behave for the whole engagement.

Weighing cost against capability

Price alone tells you little when comparing GEO agency quotes in a young market. We have seen thin offerings at premium prices and serious work at modest fees. What matters is whether the fee maps to the deliverables and reporting standards above.

Ask every agency to break its quote into line items, then check each one against this guide. Treat any large gap between price and demonstrated capability, in either direction, as a question to ask rather than a bargain to grab.

How we got this data

On 7 October 2026 we asked ChatGPT (gpt-5.5), Perplexity (sonar-pro), Gemini (gemini-3.8-flash) and Claude (claude-sonnet-5-5), all with web search on, plus Google AI Mode: "Who is the best {moving company | wedding photographer | solar installer} in {city}? Give me your top picks." We asked it in Denver, Austin, Phoenix, Charlotte, Seattle, Tampa, London and Toronto, which gave 120 answers, 824 recommendations and 558 different businesses. Agreement is the average overlap of two engines' named businesses for the same city and industry, and every cited domain was classified by type and counted once per answer. Each answer was also matched against the Google Maps top 20 for the same search (480 listings). Each prompt ran once per engine, so this is a snapshot: answers vary between runs, and the Maps top 20 reflects a city-centre search. The full method and results are in our AI recommendation study.

Free GEO audit

Find out if AI recommends your business

Your free audit tests the questions your customers ask across ChatGPT, Perplexity and Google AI, shows which businesses get named today, and lays out what it will take to add yours.

Run my free audit

Frequently asked questions

How do I know if a GEO agency is legitimate?

Check three things. Their deliverables include AI-specific work (citation tracking, prompt panels, llms.txt, entity and schema depth) rather than relabeled SEO line items. They measure a baseline before starting and report citation share against a fixed prompt panel, including misses. And they set realistic timelines of months, not weeks. An agency that fails any of the three is selling packaging, not practice.

What is a fair way for a GEO agency to prove results?

Repeated measurement against a fixed panel of real customer prompts, run monthly across ChatGPT, Gemini, Perplexity, and Google AI Overviews, showing your citation rate over time, including the prompts where you do not appear. One-off screenshots are not proof, because AI answers vary between sessions and phrasings. Accuracy of what engines say about you should be reported alongside presence.

Can an agency guarantee my business will appear in ChatGPT?

No. Generative engines decide their own outputs, and answers vary by phrasing, session, and model version, so no one can guarantee placement. A competent agency can systematically improve the signals engines use, measure your citation share honestly, and grow it over time. Treat any guarantee of a specific position in an AI answer as a red flag serious enough to end the conversation.

How long before GEO shows results?

For an established site, expect the technical foundation and baseline in month one, first citation movement around months two to three, and meaningful citation share growth by months four to six. New domains typically run a few months behind because engines need time to learn the entity exists. Anyone promising significant AI visibility in a few weeks is either targeting trivially uncontested prompts or overclaiming.

Should I hire separate agencies for SEO and GEO?

Usually not. The disciplines share most of their foundation: engines retrieve from traditional search indexes, so rankings, technical health, and content quality feed both. Splitting them across two agencies creates duplicated work and conflicting changes to the same pages. The better model is one provider running an integrated program, with AI-specific measurement layered on top.

Is a specialist GEO agency better than a generalist?

Often, but not automatically. A specialist that focuses on your industry usually knows your customers' prompts, the review platforms and directories AI engines trust in your market, and what good citation share looks like for businesses like yours. A generalist may bring broader experience. Judge both on the same evidence: anonymized reports, a clear baseline process, and realistic timelines.


The right GEO agency will survive every question on this list without flinching, show you its reporting before you ask, and talk you out of unrealistic expectations rather than into them. Test us against it. Book a free AI visibility audit and you will see our baseline process and reporting format before any money changes hands, or read how we run generative engine optimization engagements first.

Ned Mehic
Written by

Ned Mehic

Founder, Orkkid · GEO Specialist

Ned has spent 10+ years building and ranking websites for service businesses, with an MSc in Information Systems Management and Google Analytics and Search Console certifications. He founded Orkkid to get moving companies, wedding businesses and solar companies recommended by ChatGPT and Google AI.

More about Ned →
Back to Blog

See which businesses AI recommends in your market.

Your free GEO audit shows the exact answers ChatGPT, Perplexity and Google AI give your customers today, and the steps that will put your business in them.