Last updated

Design and Growth

How to choose a GEO agency when every proposal promises to get you into ChatGPT

To choose a GEO agency, test what each proposal claims against what the engines publish and what the research measured. Underneath the names the work is the same three jobs, and nobody can guarantee what ChatGPT will say.

Checking a GEO agency proposal against evidence, sources and reporting requirements

Eight sentences from the proposal

Most GEO proposals contain some version of the same eight sentences. “We guarantee you will be recommended by ChatGPT.” “Your AI visibility score is low.” “We will fix your robots.txt for AI.” The other five are further down this page.

Each of them can be tested against something public. Google and the other engines publish how they find and use pages. One large peer-reviewed study measured what changes an AI answer. If you are working out how to choose a GEO agency, test the claims against those sources before you compare the promises.

The proposals describe the same work, whatever they call it. One says generative engine optimization, one says answer engine optimization, a third says LLM SEO. Under the labels, the sound ones describe three jobs. The first is making your pages findable by the crawlers assistants rely on, and the second is making them quotable when an assistant reads them. The third is keeping the facts about you consistent on the third-party sites assistants also read, such as review platforms, directories and the publications in your field. Proposals often sell that third job as digital PR or brand mentions.

None of that is secret, even when a proposal suggests it is. The engines are fairly open about how they find and use pages. Where they are open, a proposal should agree with them. Where it disagrees, it should say why.

We sell generative engine optimization services, so a proposal of ours could be one of the documents on your desk. Put every question below to us as well. The evidence for each answer comes from the engines’ own documentation and from independent research.

This article assumes you have already decided that AI answers matter to your pipeline. If you have not, start with what your buyers ask assistants about your category. Our GEO page explains what gets a business named in those answers.

Which is better, SEO or GEO?

Neither is better, because they are not rivals. For Google, GEO is SEO. For the other assistants it is SEO plus some work off your own site. Be wary of an agency that frames it as a choice.

Google is unusually direct about its own AI features. Its guidance for site owners says there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary. A page needs to be indexed and eligible to be shown with a snippet, and the usual SEO fundamentals apply.

Its guide to optimizing for generative AI search, last updated in July 2026, goes further. From Google Search’s point of view, optimizing for generative AI search is optimizing for the search experience, “and thus still SEO”. It also lists things site owners can ignore for Google Search. Among them are llms.txt files and other special markup, chunking content into small pieces, rewriting content just for AI and seeking inauthentic mentions.

GEO still has real work in it. ChatGPT, Perplexity and Claude run their own crawlers and their own search. Assistants also lean on third-party sites to decide whom to trust. So GEO is SEO done with a few more readers in mind, plus work off your own site. It does not replace SEO, and a GEO program on a site with weak search foundations usually goes nowhere.

If you have to pick one budget line, we would fix the search foundations first, because every assistant that searches the web benefits from them. If those are already sound, GEO is the extra layer that covers the assistants Google’s rules do not.

Is SEO still worth it in 2026?

Yes, and Google says so itself. Many GEO pitches open by telling you SEO is dead. Google’s own guide says its best practices for SEO continue to be relevant. Its AI features are built on the same index, and a page has to be indexed and snippet-eligible to be used.

The kind of page that does well has changed. Google’s guide asks for unique, expert-led content that gives value beyond common knowledge. It describes commodity content as generic material that adds little insight. That is a harder bar than keyword coverage, and the research below points to the same one. Proposals that promise topical authority are describing the same bar across a subject: several pages that each answer one question in depth, which is still SEO.

So treat “SEO is dead” as a warning sign in a proposal. An agency that believes it will probably neglect the indexing, speed and page quality that every assistant depends on.

Some buyers should stop reading proposals at this point. Assistants and AI features draw heavily on search indexes, so GEO rarely works on a site with weak foundations. If pages are not indexed, are slow, or say little that is specific, fix that first. B2B SEO is where we would start. If the site itself needs rebuilding, the website redesign checklist covers what not to lose on the way.

If buyers in your category do not yet ask assistants about suppliers, paid search may produce inquiries sooner. PPC for SaaS covers how to make an ad account learn from the right leads.

Claims to test in a GEO proposal

Eight claims recur in GEO pitches, and each has one question that tests it. Four are about what the agency will do to your pages, and one study tests most of those. In that study, citing sources, adding quotations and adding statistics changed AI answers most. Keyword stuffing and a confident tone did not.

The study is by researchers at Princeton University and IIT Delhi, published at KDD 2024. It is the most rigorous independent evidence we know of, and most other claims about GEO come from agencies. The researchers built a benchmark of 10,000 queries and rewrote source pages in nine different ways. Then they measured how much of each generated answer came from each source.

What the GEO study found, on its own benchmark and on Perplexity. Source: Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024.
Change to the pageWhat happened to its visibility in AI answers
Citing sources, adding quotations, adding statisticsThe top methods. A relative improvement of 30 to 40 percent on the study’s main measure and 15 to 30 percent on its subjective measure
Improving fluency and readabilityAlso a significant improvement, of 15 to 30 percent
A more persuasive, authoritative toneNo significant improvement
Keyword stuffingLittle to no improvement on the study’s engine, and 10 percent worse than doing nothing on Perplexity
Pages ranked lower in searchBenefited more. Citing sources raised visibility by 115.1 percent for the source ranked fifth, while the top-ranked source lost 30.3 percent on average when every source was optimized

On Perplexity, a real engine with a large user base, the same methods held up. Adding quotations did best on the study’s main measure, with a 22 percent improvement. On the other measure the improvements reached 37 percent.

Three limits apply before you use these numbers in a decision. The study measured how much of an answer a source contributed. It did not measure clicks, inquiries or revenue. The researchers built the benchmark themselves and tested it on their own engine and on Perplexity as it behaved in 2023 and 2024, and engines change. The authors also found the effect varied by domain, so the study does not give one recipe for every site.

Read the study as a direction. Evidence and specifics help a page, and padding and a confident tone do not. The four claims that follow are tested against it.

“We optimize your keywords for AI” is the easiest to test. Keyword stuffing did little in the GEO study and did worse than nothing on Perplexity. Google’s guide adds a warning about extra pages made to target the many related queries its AI generates. Made mainly to manipulate rankings, they break its scaled content abuse policy. Ask what will be added to each page besides keywords.

“We will make your content more authoritative” fails the same way. A more authoritative tone showed no significant improvement in the study. Sources, figures and quotations did. Ask which of those they will add, and where the facts will come from, because you will have to stand behind every one.

“We add special markup and an llms.txt file for AI” is contradicted by Google’s guidance. Google says no special files or markup are needed to appear in its AI features. It lists llms.txt among the things Google Search site owners can ignore. Standard structured data still does its ordinary job in Search, provided it matches what the page shows, and that is technical SEO rather than a separate AI layer. Other assistants may read such a file, so it is not worthless, but one file is not a plan. Ask which change, on which page, for which assistant, and what evidence says it matters.

“We will get you onto the lists AI reads” is the only one of the four that describes real work. Third-party sources do matter. Google also warns that seeking inauthentic mentions across the web “isn’t as helpful as it might seem”. Ask which sites, how you get there, and whether any placement is paid, so you can judge what it is worth.

Table pairing four GEO proposal claims with a question to ask: AI keywords, what changes on the page; authority, which sources and facts; special markup, for which assistant; third-party mentions, which sites and what method (opens full size in a new tab)
Four claims that keep appearing in GEO proposals, and the question that shows what work sits behind each one.

The crawler question

The fifth claim is “We will fix your robots.txt for AI”. A careless agency can do real harm here. Every GEO proposal should say which AI crawlers it will let into your site, and it should treat training crawlers and search crawlers differently.

Assistants find pages through crawlers, and most AI companies now run more than one. One collects content for training models, one builds a search index, and one fetches a page when a user asks. Blocking each has a different effect, and your robots.txt decides which may read your site.

What each crawler does, from each company’s own documentation, read October 2026.
CrawlerWhat it is forWhat blocking it does
OAI-SearchBot (OpenAI)Surfaces websites in ChatGPT’s search featuresYour site will not be shown in ChatGPT search answers, though it can still appear as a navigational link
GPTBot (OpenAI)Crawls content that may be used to train OpenAI’s modelsSignals your content should not be used for training. It is not the search crawler
ChatGPT-User (OpenAI)Visits a page when a ChatGPT user’s request needs itUser-initiated, so robots.txt rules may not apply; it plays no role in search visibility
PerplexityBotSurfaces and links websites in Perplexity’s search results; not used for trainingYour site drops out of Perplexity’s results
Claude-SearchBot (Anthropic)Improves search result quality for Claude usersMay reduce your visibility and accuracy in Claude’s search results
ClaudeBot (Anthropic)Collects content that could contribute to model trainingSignals your content should be excluded from training
Google-ExtendedControls whether Google may use your content to train future Gemini models, and for grounding answers in Gemini Apps and Vertex AIGoogle says it does not affect inclusion in Google Search and is not a ranking signal. It does keep your content out of grounding in Gemini Apps

The sources are OpenAI’s crawler documentation, Perplexity’s, Anthropic’s and Google’s list of its crawlers. They change often, so check each before acting on the table.

Google adds one control that is not a crawler. Since August 31, 2026 every site has a Search generative AI control in Search Console, under Settings. It decides whether the site’s links and content can appear in AI Overviews, AI Mode and generative AI features in Discover, and inclusion is the default. Google says the setting is not a ranking or inclusion signal for the rest of Search, and it does not control AI training. Any baseline should record which way it is set.

Whether to let your content train models is a business and legal decision, and it belongs to you. Whether to let search crawlers in is a visibility decision. Blocking them removes you from the answers you are paying an agency to get you into. A competent proposal separates the two, names the crawlers it will allow, and says why. A proposal that offers to unblock every AI bot, or to block them all, has not made that separation.

So ask which crawlers they will allow and which they will block, and whether they separate training crawlers from search crawlers. If the answer is vague, keep them away from the file.

What a GEO engagement should measure

The last three claims are about results. AI visibility is harder to measure than rankings, so a proposal should say how it will measure it.

“Your AI visibility score is low” usually comes first, as the reason to buy. Ask which questions were asked, on which assistants, how many times, and whether the check can be repeated to show change. Google states that no third-party tool has access to its internal ranking or AI systems, so any score is an estimate built from sampling. An estimate is usable when the proposal shows the method. Without the method, set the score aside.

“We guarantee you will be recommended by ChatGPT” cannot be true. No agency controls what an assistant says, and the same question can get a different answer an hour later. Ask what will be measured instead, on which assistants, and how often.

“You will see results in weeks” may or may not be true. Ask for the baseline they will measure from, and what result would show the work is not working.

In place of those three, agree four measures before the work starts. The first is a fixed set of the questions your buyers ask, put to named assistants on a schedule. Each run records whether you are named, cited or linked. Keep the set the same every time so the results can be compared. Tools sold for AI visibility tracking automate this measure, so ask how a tool samples each assistant before you trust its trend line. The second is which of your pages get cited, so the work can follow what assistants use.

The third is Google’s Generative AI performance report in Search Console. Google recommends it for seeing how your content performs in its generative AI features on Search. Read it alongside ordinary search performance for the pages being changed.

The fourth, and the closest to revenue, is inquiries that arrive from AI answers. The simplest instrument is a question on your inquiry form asking how the buyer found you, with the assistants listed as answers. It is crude, and it is the measure we trust most.

Agree what would count as failure too. If the baseline questions show no change after a fair period, a good agency will say so, keep the measure and change the plan.

Five steps for measuring GEO: fixed buyer questions, named assistants, scheduled answer checks, pages cited or linked, and attributed inquiries (opens full size in a new tab)
Ask the same buyer questions of the same assistants on a schedule, so a change in citations means something.

Which GEO agencies are the best?

You will find plenty of ranked lists, and most are written by agencies that put themselves on them. We are not going to add another. The more useful question is which agency is best for your site, and you can answer that from the proposals.

The guides that rank for “how to choose a GEO agency” share the problem. Google showed eight of them on October 9, 2026, and we read the seven that opened for disclosure, sources and crawler advice. Six were published by agencies that sell GEO or SEO. The seventh is a ranked list on a CRM vendor’s blog, and it says it is “brought to you by” the agency it ranks first. Only one of the seven clearly says the author has an interest in the answer.

None of the seven links to Google’s own guidance that AI search needs no special optimization. None separates training crawlers from search crawlers in its robots.txt advice. The two pages that name AI crawlers both treat GPTBot and ClaudeBot, which collect training data, as if blocking them would stop citations. Only one cites the original GEO research paper.

We sell this work too, so weigh what we say the same way. Agencies can write usefully about choosing an agency. The guides you will find mostly skip the checks on this page, and those checks are how you tell one proposal from another.

The proposal worth buying usually has these qualities, and the weak ones usually lack most of them:

  • It starts with a baseline: the questions your buyers ask, put to named assistants, with the results recorded.
  • It names the pages it will change and what it will add to each, in terms that match the research: sources, figures, quotable answers.
  • It separates training crawlers from search crawlers in its robots.txt advice.
  • It covers the third-party sites assistants cite for your category, and says how it will get you onto them, including whether any placement is paid.
  • It makes no guarantee about what an assistant will say.
  • It says how often visibility will be re-checked, and how.
  • It connects the work to inquiries, not only to visibility.
  • It leaves everything it builds in your accounts and on your site.

The general rules for picking any marketing agency still apply on top. Meet the people who would do the work, not only the people who sell it. Ask for work in a field like yours and for someone you can speak to about it. Read the reviews on a platform the agency does not control, including the critical ones. And check that the contract lets you leave with your data and your access intact.

If the proposals in front of you are hard to tell apart, send them to us with your site address. You will have a reply within 12 hours. The people who would do the work will then spend 30 minutes on a call with you. They will go through the claims in each proposal, ours included if we pitch, and tell you whether the work you need is GEO at all.

Questions about choosing a GEO agency

How do I choose a GEO agency?

Ask each agency for the baseline it will measure from, the pages it will change and what it will add to each, how it treats AI crawlers, which third-party sources it will work on, and how it connects the work to inquiries. Then test its claims against the evidence. Guarantees, secret markup and keyword work are the warning signs.

Which is better, SEO or GEO?

Neither replaces the other. For Google, its own guide says optimizing for generative AI search is “still SEO”, with no special markup or files needed. GEO adds work for assistants that run their own crawlers, such as ChatGPT, Perplexity and Claude, and work on the third-party sites they read, and it rarely works on a site with weak search foundations.

Is SEO still worth it in 2026?

Yes. Google says its SEO best practices continue to be relevant to its AI features, which need a page to be indexed and eligible for a snippet. What rewards a page has shifted toward unique, expert-led content that goes beyond common knowledge.

Which are the best GEO agencies?

Most published lists are written by agencies that rank themselves, so judge the proposal in front of you instead. The best one for you starts with a measured baseline, names the pages and changes, separates training and search crawlers, makes no guarantees, and ties the work to inquiries.

Can a GEO agency guarantee my company will appear in ChatGPT?

No. Nobody controls what an assistant says. An agency can make your pages crawlable, quotable and corroborated, and measure how often you are named for a fixed set of questions. A guarantee beyond that is a sales line.

Should I block AI crawlers?

Separate the two kinds. Blocking training crawlers such as GPTBot or ClaudeBot keeps your content out of model training and, by their owners’ accounts, does not remove you from search. Blocking search crawlers such as OAI-SearchBot, PerplexityBot or Claude-SearchBot removes or reduces you in those assistants’ answers.

What is the difference between a GEO agency and an SEO agency?

Most of the work is shared. Traditional SEO covers technical SEO, content and links, and Google says that is also what its AI features use. A GEO agency adds work for the assistants that run their own search, such as ChatGPT, Perplexity and Claude: crawler policy, the third-party sites they read, and visibility measured in their answers. An SEO agency that already does that extra work is doing GEO under another name.

The author

Rajdeepsinh Jadeja

Head of Marketing · 10+ years in software

Rajdeepsinh Jadeja is Head of Marketing at Redlio Labs, where he has worked since 2016. He runs Google Ads, Meta Ads and SEO, including AI search visibility, for Redlio Labs and its clients in the US, UK, Australia and the UAE. He writes the marketing articles on this site.

  • Google Ads
  • Meta Ads
  • SEO
  • AI search (GEO and AEO)
  • Conversion tracking
Full profile

Keep reading

Start with a conversation, not a contract

Bring us the system to build, the product to design, the inquiries to grow, or the team to extend.

Prefer email?info@redliolabs.com

We use what you send to understand and answer your inquiry. Privacy policy

  • We reply within 12 hours
  • A quote within 48 hours of the scoping call
  • Direct access to the delivery team
  • Peer-reviewed releases
  • NDA before your first call
Mayursinh Jadeja

Mayursinh Jadeja

Founder, Redlio Labs

Rated 5.0 on Clutch and 5.0 on Google