[AI Search Visibility]
How to Brief a GEO Agency: 12 Questions to Ask Before You Sign
Most GEO pitches sound the same until you ask the right questions. Here's the 12-question brief that separates a real citation engineering operation from a rebranded SEO retainer.
The short answer
A strong GEO agency brief forces the agency to answer specifics before you sign: which engines they track, how they define citation share, what proof they can show with dates, and what happens contractually if citations don't move. Run the 12-question checklist below, each with a green flag and a red flag, before you commit a budget.
What is a GEO agency briefing, and why do you need one before the first call?
A GEO agency briefing is a short document you send before any sales call: what you sell, who buys it, which AI engines matter to your category, and the 12 questions you plan to ask. You need one because generative engine optimization is young enough that agencies sell wildly different services under the same label, from a single tweak aimed at Google's AI Overviews to a full content operation publishing across every major engine.
The term itself has a real, dated origin. A November 2023 paper on generative engine optimization, co-authored by researchers including Pranjal Aggarwal, showed that deliberate content optimization could lift a brand's visibility in generative engine responses by up to 40 percent (arXiv, 2023). That paper is also why the category attracted so many agencies so fast: a documented 40 percent lift is a good pitch, and not every agency selling it can back it up.
GEO also gets confused with answer engine optimization, a related but narrower practice focused on structuring content so it gets pulled into direct-answer boxes and featured snippets. A brief that makes an agency define both terms in its own words, on the spot, tells you more than any deck they bring.
Treat the brief as a filter, not a formality. Send it to every agency on your shortlist before the first call, and score the answers against the checklist below. The agencies that answer in specifics move to a second call. The ones that answer in adjectives don't.
What should you ask about their methodology and engine coverage?
Ask which engines they track and optimize for, by name, because 'AI search' is not a methodology. ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews are separate retrieval systems with separate citation behavior, and an agency that only talks about 'ranking in AI' is describing one of them at best.
The gap between classic SEO and citation work is measurable. Ahrefs' analysis of AI-cited URLs found that only 12 percent of the URLs cited by AI assistants also rank in Google's top 10 for the same prompt (Ahrefs, 2025-08-11). An agency still optimizing for classic rankings is optimizing for the wrong list roughly seven times out of eight. Ask them to walk through how their process differs for a ChatGPT citation versus a page-one Google ranking, and listen for a real answer, not a rebranded SEO checklist.
A credible agency should also be blunt about what it will not do. The mechanism behind durable citations is quality, coverage, and freshness at a scale most in-house teams cannot match, not a trick that games the model. If anyone on the call implies they can manipulate what an LLM cites, that is the fastest disqualifier on this whole list.
Also ask how they think about freshness. AI engines tend to favor recently updated, well-sourced pages over stale ones, so a methodology that treats an article as 'done' after one publish date is missing half of how citation work actually functions.
| # | Question to ask | Green flag | Red flag |
|---|---|---|---|
| 1 | How do you define GEO and AEO for our category? | Gives a precise definition tied to citation share and names specific engines | Recycles generic SEO language or claims to hack the models |
| 2 | Which engines do you track and optimize for? | Names ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews explicitly | Talks only about 'AI search' or Google AI Overviews alone |
| 3 | What is your headline metric and how is it calculated? | Defines citation share, answer presence, and share of voice with a real formula | Reports only rankings or organic traffic |
| 4 | Can you show a case study with a named client, dated numbers, and sign-off? | Named client, specific timeframe, offers a reference call | Anonymized 'leading company,' no dates, no way to verify |
| 5 | What is your realistic publishing cadence? | Gives a concrete weekly or monthly number and how it scales | Vague answer like 'as much as needed' |
| 6 | Who owns the content and domain equity once we leave? | Content and rights transfer fully to you, published on your own domain | Agency-owned microsite or reluctance to answer |
| 7 | How do you source and verify statistics in what you publish? | Describes a real fact-checking step and requires a dated source per claim | Can't describe a sourcing process |
| 8 | What's in the report, and how often do we get it? | Recurring report with citation share by engine and the prompts tested | 'We'll reach out if something interesting happens' |
| 9 | What's the realistic range before we see first citations? | Gives a range grounded in past engagements, no fixed date promised | Promises a specific citation count or date |
| 10 | What happens if citation share doesn't move? | Defined revision process, ramp period, and recourse in writing | No contingency, results discussed only if you raise it |
| 11 | Do you ever try to manipulate or game what a model cites? | Explicit no, explains reliance on quality, coverage, and freshness | Any hedge, or a described manipulation tactic |
| 12 | What's the contract length and exit clause? | Month-to-month or short term with a clear notice period | 12-month-plus lock-in with no early exit |
What proof should you demand before you believe a case study?
Demand a named client, a specific timeframe, and a real number, not an anonymized 'leading SaaS company saw incredible growth.' A credible GEO case study reads like a lab result: starting point, end point, dates, and what changed in between.
That proof matters more than it used to because the sales cycle now starts before your team ever hears about it. A multi-source analysis covering 680 million AI citations found that 73 percent of B2B buyers now use AI tools such as ChatGPT and Perplexity during purchase research (PR Newswire / Loganix, 2026-04-03). The case study an agency shows you is a preview of what is about to happen to your own prospects: they are asking an AI which vendor to trust before they ever fill out your form.
Ask, too, whether the client has signed off on the numbers being shared with you. A serious agency treats client data the way a serious publication treats sources: nothing goes out named until the client has confirmed it. If an agency can't tell you whether a case study is cleared for use, don't take the numbers at face value.
If they can't produce a single verifiable example on the call, that's the answer. Move on.
How should a GEO agency measure and report citation share?
A GEO agency should measure citation share, the percentage of relevant AI answers in your category that cite you, and be able to define it precisely, not gesture at 'AI visibility.' Push for the full metric stack: citation share as the headline number, citation count per day for volume, answer presence for how much of the question universe you cover, and share of voice against named competitors.
Ask what the report actually contains before you sign anything. A report that lists citation share by engine, week over week, with the specific prompts tested, is a different product from a monthly email that says things are 'trending up.' If they can't show you a sample report on the call, ask them to send one before the next one.
Ask, too, how citation share compares to what you're already tracking in organic rankings or paid spend. An agency that can put citation share next to your existing KPIs, on the same dashboard, is treating this as a real discipline. One that can only talk about it in isolation is still figuring out how to prove its own value.
What should you ask about content production and who owns it?
Ask for their realistic publishing cadence and get a number, not a range like 'a lot.' Citation share moves with coverage and freshness, and an agency that ships a handful of pieces a month is playing a different game than one running a genuine content operation. Ask what a typical week of output looks like and whether it scales as your question universe grows.
Then ask who owns it. Content, and the domain equity it builds, should belong to you, published on your own site, not parked on an agency-owned microsite you lose access to if you leave. This is a five-minute question with a permanent consequence, so get the answer in writing before you discuss article terms.
Also ask how they source and verify statistics in what they publish under your name. Every claim in a citation-grade article needs a real, dated source. An agency that can't describe a fact-checking step, or that shrugs when you ask where a number came from, is one bad statistic away from becoming the reason an AI answer stops trusting your domain.
Ask what a 'question universe' means for your category specifically: which comparison prompts, 'best of' queries, and buyer questions they're mapping content against. A vague content calendar is not a citation strategy.
What contract terms protect you if citations don't move?
Ask what happens if citation share doesn't move, and get the answer before you sign, not after the first disappointing report. A fair contract has a revision process, a defined ramp period, and language about what recourse looks like if the work isn't landing.
Be wary of anyone who guarantees a specific citation count or a ranking by a fixed date. Nobody controls what a model decides to cite, including the agency you're hiring. What a credible agency can control, and should commit to in writing, is coverage, cadence, freshness, and how it reports progress against those inputs.
Check the term length and the exit clause. Month-to-month or a short initial term with a clear notice period tells you the agency expects to earn the renewal. A 12-month lock-in with no early exit tells you they expect you to stay for reasons other than results.
Ask about the ramp period explicitly: most citation work takes time to accumulate as new content gets indexed and cited, and a credible agency will tell you what a normal early curve looks like rather than promising an immediate spike.
What are the clearest red flags in a GEO agency pitch?
The clearest red flag is a guarantee: guaranteed citations, guaranteed rankings, or a guaranteed date. Nobody can promise what a language model will choose to cite, and anyone who does is selling confidence, not a methodology.
- Refuses to name the specific engines it tracks (ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews) and talks only about 'AI search' in general
- Case studies with no client name, no timeframe, and no way to verify the numbers
- Can't define citation share, answer presence, or share of voice without stalling
- Any mention of manipulating, gaming, or tricking the model, including hidden text or fabricated citations
- Vague or infrequent reporting with no visibility into which prompts were tested
- Long contract lock-in with no exit clause and no revision process if results stall
Key takeaways
- A GEO agency brief should force specifics: named engines, a defined metric stack, and a dated case study, not a generic AI-search pitch.
- Only 12 percent of URLs cited by AI assistants also rank in Google's top 10 for the same prompt, so an agency still optimizing for classic rankings is solving the wrong problem (Ahrefs, 2025-08-11).
- 73 percent of B2B buyers now research vendors using AI tools before ever talking to a rep, which is why the case study an agency shows you matters (PR Newswire / Loganix, 2026-04-03).
- Any promise to manipulate, hack, or game what a model cites is an immediate disqualifier, not a selling point.
- Content and domain equity should transfer to you, not stay parked on an agency-owned property.
- Guarantees on citation counts or ranking dates are a red flag; commitments on coverage, cadence, and reporting are not.
Omnicite Editorial. "How to Hire a GEO Agency: 12 Questions to Ask" The Citation Report, Omnicite. https://omnicite.co/blog/how-to-brief-a-geo-agency/
Sources
73% of B2B buyers use AI tools such as ChatGPT and Perplexity during purchase research PR Newswire / Loganix, 2026-04-03
Only 12% of URLs cited by AI assistants also rank in Google's top 10 for the same prompt Ahrefs, 2025-08-11
Deliberate content optimization can lift a brand's visibility in generative engine responses by up to 40%, the original research that coined the term generative engine optimization (GEO) arXiv (Aggarwal et al.), 2023-11-16
Frequently asked questions
What does GEO stand for?
GEO stands for generative engine optimization, the practice of structuring and publishing content so AI systems like ChatGPT, Perplexity, and Gemini cite it in generated answers.
Is GEO the same as AEO?
No. GEO and answer engine optimization overlap but aren't identical: GEO targets citations inside generative answers across multiple engines, while AEO focuses more narrowly on structuring content for direct-answer boxes and featured snippets.
How long does it take to see AI citations after hiring a GEO agency?
There's no universal timeline, and any agency that gives you a guaranteed date is overpromising. Ask instead for the range they've seen across past engagements and what inputs (coverage, cadence, freshness) drive that range.
How much should a GEO agency cost?
Cost varies by scope, from a one-time audit to a full content operation publishing daily across engines. Price against deliverables, coverage, and reporting, not hours billed.
Can I do GEO in-house instead of hiring an agency?
You can, if your team can sustain the fact-checking rigor and publishing cadence citations require. Most in-house teams struggle less with strategy than with volume, since the gap is usually the scale needed to build coverage across dozens of category questions.
What's the single most important question on the checklist?
Whether the agency ever tries to manipulate or game what a model cites. Everything else on the list is negotiable technique; this one is a boundary, and the answer needs to be an immediate no.