How to Tell if a GEO Agency Is Faking It: 7 Questions to Ask

Can they show you a live citation, not a screenshot?

The first test is the simplest: ask the agency to pull up a real AI answer, right now, that names a client or a comparable brand in the response text, with a source link you can click and verify. A real GEO practitioner does this by opening ChatGPT, Perplexity, or Google AI Overviews and running the actual prompt, live, on the call. An agency that instead opens a slide deck with a static screenshot, or a dashboard showing a proprietary 'AI visibility score' with no underlying answer attached, is showing you a proxy metric it built, not evidence of a citation.

This matters because citations are the only outcome that counts. Rankings and traffic are SEO's currency; a GEO agency that cannot produce a live, checkable citation is measuring the wrong thing, or has not actually placed you anywhere yet.

Do they know the difference between retrieved and cited?

Retrieved means a model's search layer pulled your page into its context window while answering a prompt. Cited means the model actually named you or linked your page in the final answer text. These are different events, and an agency that treats them as the same thing is either new to this work or hoping you will not ask. A domain can show up in a model's crawl logs constantly and never once appear in an answer a buyer reads.

Ask directly: of the AI answers you are tracking for us, what share cite us versus just retrieve us. A competent agency has this number, or can get it within a reporting cycle. One that answers with 'we're seeing strong visibility' without naming a citation rate is restating a vibe, not a metric. What a real AI visibility audit should measure covers this gap in more depth.

Do they report citations, or just traffic and rankings?

The right report shows which AI engines named you, for which prompts, how often, and what the source page was. That is a different report than organic sessions, keyword position, or a generic 'AI overview appearances' count with no prompt-level detail behind it. If the monthly report you get looks identical to the SEO report you got in 2024 with a new logo on the cover, the underlying work probably has not changed either.

Ask to see the actual prompt list they track and how the citation rate for each one moved month over month. For a plain-language walkthrough of what to look for in that report, see how to measure AI search visibility.

Do they push Reddit seeding or bought upvotes?

If an agency's plan for winning citations from Google AI Overviews or ChatGPT includes seeding Reddit threads with planted accounts or buying upvotes to push a thread up the rankings, walk away. That is astroturfing, and The Verge has documented the spam it is already producing across Reddit as brands chase AI citations. The platforms and the model providers are actively working to detect and discount exactly this kind of manufactured activity.

The durable version of community credibility is earned: real participation, real usefulness, real third parties saying something true about you unprompted. That is slower to build and cannot be faked convincingly for long, which is precisely why it is the version that survives a platform's next detection pass.

Do they name which engines they track, and how those engines differ?

ChatGPT, Google AI Overviews, Gemini, and Perplexity do not source their answers the same way, and an agency that talks about 'AI search' as one undifferentiated channel has not looked closely at any of them. Ask which specific engines they monitor for you, and ask them to describe, in specific terms, how one engine's sourcing behavior differs from another's on your category of prompts.

A vague answer like 'we track visibility across all major AI platforms' is not a differentiated answer. A specific one names the engines, describes at least one behavioral difference between them, and explains how that difference changes what they build for you.

Do they guarantee AI citations?

No legitimate agency can guarantee a specific citation in ChatGPT, Gemini, or Google AI Overviews, because neither the agency nor the client controls how the model synthesizes its answer on any given prompt. What a real agency can commit to is the input side: the corroborating sources, the structured content, and the earned mentions that make a citation more likely over time, tracked against a citation rate that should improve.

Treat any flat guarantee, one like 'we will get you cited in ChatGPT within 30 days' with no qualifiers, as a sales tactic borrowed from an SEO pitch that promised page-one rankings. The mechanism is different, and the promise does not transfer.

Can they point to earned corroboration, not just their own content?

Ask for examples of third-party pages, meaning press coverage, review sites, comparison posts, forum threads, that mention a client and that the agency did not write or publish itself. Models weight independent corroboration differently than a brand's own marketing, so an agency whose entire body of work is blog posts on the client's own domain is building half a strategy.

This is also where you can judge whether an agency operates as a genuine digital agency that builds B2B authority or just produces content. Executive thought leadership, earned press, and community standing compound because other sources point back to them; a stack of unlinked blog posts does not. Corbelix treats corroboration as the load-bearing part of the work, which is also why we recommend running your own check with the AI visibility grader before you take any agency's word for your current standing.

The seven questions, side by side

Keep this table in front of you on the call. A vendor that lands in the red flag column on two or more of these questions has not built a real citation practice yet, whatever the pitch deck claims, and no single strong answer offsets multiple weak ones elsewhere in the table.