How to Brief a GEO Agency: A Template for Marketing Managers

A strong GEO agency brief names a business objective, a dated AI visibility baseline, a fixed prompt panel, a clear split of responsibilities, and a reporting format you can verify without the agency’s cooperation. Most disappointing engagements trace back to a thin brief, not a weak agency. This guide walks marketing managers through each part of a **GEO agency brief template**, from internal groundwork to the mistakes that derail results, so you can brief any generative engine optimization agency with confidence.

What internal groundwork should you finish before writing the brief?

Finish four things before you write a word: assign owners, confirm site readiness, settle account ownership, and set your crawler and AI policy. Agencies can only deliver against the workstreams your team staffs, so unowned work is where most engagements stall. A common failure is scope collision, where the agency assumes someone on the client side owns a workstream that nobody actually covers.

Site readiness comes first. Google’s May 2026 guide to AI features in Search, covered by Search Engine Journal, frames this work as “still SEO”. Indexability, snippet eligibility, and a crawlable site remain the entry requirements, and Google says llms.txt and special AI schema are not needed for its own generative features. Check these basics before you ask an agency to build on top of them.

Account ownership also needs settling early. Google Analytics 4, Google Search Console, Bing Webmaster Tools, and tool subscriptions should sit in your company’s name, with the agency added as a user. If the agency creates them under its own login, you lose your history when the engagement ends.

Finally, write down your crawler policy. Training crawlers such as GPTBot, ClaudeBot, and Google-Extended are now separate from AI search crawlers such as OAI-SearchBot, Claude-SearchBot, and PerplexityBot. A brand can block training use while staying eligible for citations in AI answers. State your position in the brief and make sure your CDN and WAF settings match your robots.txt file. Add your position on Search Console’s generative AI opt-out control at the same time.

How do you set business goals and GEO objectives?

Start with the business outcome, then translate it into a measurable end state you can verify without the agency’s help. A weak brief asks for activity, such as “optimize our site for AI search”. A strong brief names a result, such as closing the gap with a named competitor on integration and comparison prompts by a set date.

Open the brief with your commercial objective and your current search ecosystem. Then list the capabilities you expect the agency to bring: AI visibility research, entity and authority strategy, answer-ready content, technical implementation, and performance measurement. This five-part structure mirrors how agency RFP guides, including one from Gigawatt Group, frame the evaluation.

Ask for realistic targets, not guarantees. No contract can honestly promise placement inside an AI answer, because agencies influence inputs such as content, authority, and structure, while the AI systems control the outputs. A good agency commits to deliverables, work, and reporting.

Set expectations on traffic as well. AI visibility is not the same as AI traffic. Statcounter put AI chatbots at roughly 0.25% of tracked referrals in August 2026, so a successful GEO program may show up first as citations and brand mentions, then later as pipeline.

What AI visibility baseline should you share with an agency?

Share a dated baseline made of a prompt set, the engines it ran on, saved answers, and a list of the brands and URLs that appeared. If you have none, ask the agency to produce it before changing anything. A serious agency measures first. A weak answer is a proprietary “score” with no prompt list behind it.

A baseline that fits a brief usually contains 10 to 30 high-value prompts, a report on mentions, citations, positioning, and which competitors win each prompt, and a prioritized fix plan. Raw records matter: prompt text, full answers, timestamps, engine, cited URLs, and repeated runs. Require that the agency hand all of it to you.

Repeated runs matter because AI answers vary. A SparkToro and Gumshoe.ai study, reported by Search Engine Journal, found that ChatGPT, Claude, and Google AI almost never return the same brand list twice. How often a brand appears across many runs is a steadier signal than any single answer or its position in a list.

Run your own spot checks too, with search switched on in ChatGPT, Perplexity, Gemini, and Claude, and ask each question in a fresh conversation. Without search on, you test model memory, not access to current sources. For ongoing tracking, tools such as Profound, Peec AI, Semrush’s AI Visibility Toolkit, Ahrefs Brand Radar, and Otterly.ai cover this ground. Verify pricing on vendor pages, since it changes often. When you evaluate a tracker, check engine coverage, whether it separates mentions from citations, who controls the prompts, and whether it keeps week-over-week history.

How do you define target prompts, topics, and competitors?

Define a fixed prompt panel built from real buyer questions, name the engines to monitor, and list the competitors you will be measured against. Sources disagree on size. SE Ranking suggests starting with 20 to 40 prompts across two or three models and tracking them for at least 30 days, while larger B2B programs run into the hundreds.

Build the prompt set from buyer decisions

Cover distinct buyer decisions, not paraphrases of one intent: category discovery, problems, comparisons, products, and brand accuracy. A practical split for a 20 to 40 prompt set puts the most prompts in consideration, then awareness, then brand evaluation, with branded prompts tracked separately from category prompts. Branded-only tracking flatters results, because buyers search for the problem, not your name.

Pull prompt ideas from paid search data, sales call transcripts, customer interviews, and support tickets. Involve sales and product teams in designing the set, since they hear the questions buyers actually ask.

Fix the engines, competitors, and topic scope

Name the engines in the brief. Perplexity almost always cites sources, ChatGPT less often, and tracking ChatGPT alone is too narrow. Share of voice only means something against a defined competitor set on the same prompt set, so list those competitors by name and keep the panel stable.

Include third-party source types in the topic scope. Peec AI’s analysis of 30 million cited sources, reported by Search Engine Land, ranked Reddit first across several engines, with YouTube, LinkedIn, and Wikipedia also in the top five. Other studies rank domains differently by engine and method, but they agree that third-party sources carry much of the weight.

What scope, deliverables, and responsibilities belong in the brief?

List in-scope page templates, off-site work, content production, approvals, and contract terms in plain language. Replace “the site” with named templates such as product, category, location, article, service, and FAQ pages. Replace “ongoing optimization” with deliverables you could tick off.

Cover off-site work explicitly. Citability is largely built outside your own domain, through reviews, PR, Reddit, YouTube, and industry lists. A plan limited to your own blog is half a plan. If the whole plan fits on your own site, it is arguably SEO under a new name. Specify whether third-party placement budget sits inside the retainer or is billed on top.

Clarify who writes the content and how subject-matter expertise leaves your team. Original insight is what gets cited, so a named expert should review new content for factual accuracy. Add regulated claims, legal review, outdated facts, inconsistent executive bios, negative sentiment, and AI hallucinations to the scope as well.

Use a RACI-style split with named approvers across content, technical, PR, and product teams. Then check the contract against six areas: scope, term and exit, content ownership, account ownership, reporting, and what it says about results. Reject clauses that return content ownership to the agency on termination, and require explicit consent before the agency uses your work in case studies. Pricing varies widely. Published GEO retainers commonly start around €3,000 per month, and whether PR or placement is included shifts the total. Many companies begin with a three-month pilot to test an agency’s capability before committing further.

Which reporting cadence, KPIs, and benchmarks should you require?

Require weekly tracking, monthly reporting, and quarterly deep dives, built around a short list of KPIs with agreed definitions. Weekly swings are noise at the executive level, so wait for a trend across at least four weekly checks before drawing conclusions.

Core GEO KPIs recur across sources, though names differ: citation rate, AI share of voice, AI referral traffic, conversion value, and sentiment and accuracy. Share of voice is your mentions or citations divided by the total across all measured brands on the same prompt set. Separate mentions from citations, and owned citations (your own domain) from earned citations (third-party pages that mention you).

For traffic, GA4 added a native “AI Assistant” default channel in May 2026. Its documentation names ChatGPT, Gemini, DeepSeek, Copilot, and Grok, but not Perplexity, and sessions without a referrer are excluded. A custom channel group placed above Referral gives fuller coverage. Confirm the details in Google’s own documentation.

Ask for a report that opens with business context, then shows what changed in the period, real citation examples, AI referral traffic, and next actions. The agency should report declines as well as wins, since some prompts will always drop. Be honest about attribution too. An agency cannot fully separate GEO effects from SEO effects, so measure citations and traffic separately, and treat vendor benchmarks as illustrative targets, not promises.

What does a copy-and-paste GEO agency brief template look like?

A GEO agency brief template has ten fields: business need, goals, baseline, prompts, scope, responsibilities, reporting, policies, contract terms, and proposal requirements. Copy the structure below and replace the bracketed text.

  1. Business need and non-goals: [Commercial objective, what you do not want the agency to touch, constraints, and named approvers.]
  2. Measurable objectives: [Example: close the citation gap with [Competitor A] on comparison prompts by [date]. Targets, not guarantees.]
  3. Baseline: [Existing prompt records, or a request for a dated baseline file with prompts, engines, raw answers, timestamps, cited URLs, and repeated runs. Four weeks of frozen measurement before any reporting.]
  4. Prompt panel: [20 to 40 prompts split into awareness, consideration, and brand evaluation, including non-branded category prompts. State who chooses the prompts.]
  5. Engines and competitors: [ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and AI Mode. Named competitor panel.]
  6. Scope: [In-scope page templates, off-site work (reviews, PR, Reddit, YouTube), content production, SME review process, and who covers placement budget.]
  7. Responsibilities: [RACI table across content, technical, PR, and product, with approval rights.]
  8. Policies: [Crawler policy, Search Console AI opt-out position, regulated claims and legal review rules.]
  9. Reporting: [Weekly tracking, monthly executive report, quarterly review. Required KPIs, raw data access, and a change log for any site access.]
  10. Contract terms: [Ownership of content and accounts, exit terms, pilot length, consent for case studies.]

Add questions that agencies must answer in their proposal. Which engines do you monitor and how often? What do you do beyond the website? How do you separate GEO and SEO effects? Can you show a live AI answer that cites a client page you wrote? Does your advice align with Google’s official guidance on generative AI features? Adapt further questions from published RFP guides such as Seer Interactive’s GEO RFP guide.

If you prefer to run part of this in-house, the WP SEO Agent inside WordPress handles keyword research, GEO-ready content, audits, and tracking across Google and generative engines, while WP SEO AI specialists refine strategy and step in where human judgment matters.

What common briefing mistakes derail GEO results?

The most damaging mistakes are accepting guarantees, skipping the baseline, tracking one engine, and treating GEO as a one-time project. Each one weakens your ability to judge whether the agency is working.

  • Accepting guaranteed citations or rankings. Agencies influence inputs. They do not control AI outputs.
  • Skipping the baseline. Without dated starting data, every later improvement is a claim you cannot verify.
  • Tracking a single platform or only branded prompts. Results look good on paper while real buyers see competitors.
  • Accepting relabelled SEO. Reports on Google rankings and organic traffic are not GEO reporting. Ask for a citation report.
  • Allowing undisclosed scoring. A visibility score with no stated method or prompt list cannot be audited.
  • Treating GEO as a one-time audit. Visibility drops when nobody re-tests prompts or refreshes pages.
  • Over-relying on schema or llms.txt. Google says neither is needed for its features. Official guidance from OpenAI, Anthropic, and Perplexity on this is thin, so avoid claims either way.
  • Allowing manipulative tactics. Fake reviews, bought mentions, and Reddit manipulation create liabilities that land on your brand.
  • Signing a long contract before trial work. A short pilot protects you.
  • Granting open admin access with no change log. Every site change should be traceable.

Google also states that you remain responsible for the companies you hire, so a thorough brief doubles as risk management. Freeze your prompt set, date your baseline file, and send the brief to two or three agencies. The proposals that answer your template line by line will show you who is ready to work.

This content was generated with the help of AI and it may contain mistakes

Dive deeper in

September 4, 2026

Not all AI search visibility dashboards are equal. Learn the five metrics that actually drive decisions....

July 12, 2026

Leicester SMBs are losing leads to AI search. Here's how GEO fixes that invisibility problem fast....

May 20, 2025

SEO, or Search Engine Optimization, is the process of improving your website to increase its visibility when people search for...

August 8, 2026

Most businesses misread the EU AI Act — here are 10 dangerous myths debunked with facts that matter now....

Your customers are asking AI. We make sure you're the answer.

In a quick demo, we show how WP SEO AI tracks your AI visibility, finds content gaps, and helps your company show up in ChatGPT, Google AI Overviews and more.