We asked ChatGPT and Perplexity for the best GEO agencies. They barely agreed.

If a buyer asks an AI assistant to name the best GEO or AEO agency, which companies come back? We ran the test on 31 October 2026. We asked ChatGPT and Perplexity the same kind of question a buyer would, and looked at what Google surfaces for it. The short version: the engines almost never named the same companies, and a lot of what they read is agencies ranking themselves.

How we tested it

No tricks. We asked plain buyer questions in a normal browser:

• "best GEO (generative engine optimization) agencies"
• "best AEO (answer engine optimization) agencies"
• "which agencies help brands get recommended and cited by ChatGPT"
• "best GEO and AEO agencies to get my brand cited by AI"

We ran these in Perplexity and in ChatGPT (with web search on), and we looked at Google's results for the same queries. We recorded the company names each one returned. Results move over time and by phrasing, so treat this as a snapshot, not a league table.

What each engine named

Perplexity gave a different shortlist for almost every phrasing:

ChatGPT, asked for the best GEO/AEO agencies with web search on, named a different set again, split into India-based and international:

Google, for the same searches, mostly returned "best agencies" listicles: Thrive, iPullRank, Siege Media, First Page Sage, Ten Speed, SEO Discovery, PageTraffic, Searchbloom, Digital Elevator, GrowPad and more.

Finding 1: the engines barely agreed

Put the two AI engines side by side and the overlap is tiny. Only iPullRank, Siege Media and Omniscient Digital appeared on both. Perplexity leaned toward established US content and PR agencies; ChatGPT surfaced a mix of Indian specialists and US names that Perplexity never mentioned. If you were a buyer, the "best agency" would depend entirely on which assistant you happened to open.

Finding 2: change the words, change the list

On Perplexity, "best GEO agencies" and "which agencies get you recommended by ChatGPT" produced two completely separate lists, with zero companies in common. One phrasing pulled established SEO agencies; the other pulled a newer set of AI-visibility specialists. There is no single ranking to win. You have to own a category of questions, not one keyword.

Finding 3: most "best agencies" lists rank their own author first

This is the uncomfortable part. A large share of the pages Google and the AI engines read are listicles published by agencies that rank themselves at the top. In our snapshot, Thrive's list put Thrive first (98/100), Ten Speed's list put Ten Speed first, and GrowPad, DigiMark and Digital Elevator each led their own lists. The AI engines are aware of this: ChatGPT noted that one India list was published by the agency it ranked first, so "its first-place position isn't independent validation," and Perplexity flagged self-published evidence too.

Self-ranked listicles are cheap to make, which is why everyone makes them. As more of them appear, both readers and AI tools discount them. Treat them as table stakes, not an edge.

Finding 4: the agencies that kept showing up share the same traits

The three names that recurred across engines, iPullRank, Siege Media and Omniscient Digital, did not get there by publishing more blog posts. They share the signals AI trusts:

• Independent mentions. They get named in other people's roundups, not just their own.
• Original research and data. Named frameworks, benchmarks and studies that others cite.
• Documented case studies with specifics an AI can quote (named clients, before-and-after metrics).
• A consistent identity. The same category, services and claims everywhere a model looks.

Finding 5: every engine said the same thing about guarantees

Both ChatGPT and Perplexity added, unprompted, that no agency can guarantee a ChatGPT recommendation, because answers vary by prompt, geography, context and time. ChatGPT's own buying advice was blunt: don't pay an agency just to publish 30 blogs, 20 backlinks and 10 press releases, because that is not necessarily GEO. Anyone promising "#1 in ChatGPT" is overselling.

What this means if you want to get recommended

The test lines up with how these systems actually work. There are three jobs, and they are not the same:

• Discoverability — can the AI crawlers reach and read you? (Open access for GPTBot, OAI-SearchBot, PerplexityBot and the rest; clean, server-rendered pages; schema; an llms.txt.)
• Authority — do independent sources establish you? This is where most companies are weakest. Your own site is treated as biased, so third-party mentions, reviews, directories and roundups do the heavy lifting.
• Fit — do you genuinely match the question, with evidence an AI can quote?

Publishing more of your own content helps discoverability and ranking, but it will not, on its own, get you recommended. The levers that move the needle are mostly off your own site: get named by others, publish original data, keep your brand entity consistent, and make your claims easy to quote.

See where you stand today

The fastest way to act on this is to see where you actually appear right now across Google and the AI engines. Explore AI Visibility, book a 30-minute call at cal.com/edthi-meet/30min, or message us on WhatsApp at +31 6 47553735. We'll show you your baseline and the plan to become the answer, and we won't promise you a number we can't control.