5W RESEARCH · BUYER GUIDE · UPDATED NOVEMBER 2026
A PR agency RFP works best when it scores six capabilities at fixed weights: strategy, earned media, AI visibility, integrated execution, crisis readiness and measurement. Because AI engines such as ChatGPT, Claude, Perplexity and Gemini now answer questions about which agency to hire, the RFP should also test how each finalist is described and cited in those answers, using the same prompts for every agency.
Summary: what does this guide give you?
- A weighted scorecard for six agency capabilities, with the evidence to request for each.
- A repeatable test for checking how AI engines describe each finalist agency.
- Ten questions to put to every agency, a red-flag table and a six-week process timeline.
- What should a PR agency RFP evaluate?
- How should you weight the scorecard?
- Why do AI answers matter before the RFP?
- How do you test a finalist's AI visibility?
- Which questions should you ask every agency?
- What should a proposal include?
- How do you test crisis readiness?
- What are the red flags in agency responses?
- How long should the RFP process take?
- How do you measure the agency after hiring?
- Frequently asked questions
What should a PR agency RFP evaluate?
A PR agency RFP should evaluate six capabilities: strategic fit, earned media, AI visibility, integrated execution, crisis readiness and measurement. Each one answers a different buying question, and an agency that is strong in only two or three will leave gaps the brand has to fill later.
- Strategic fit covers category knowledge, senior counsel, business understanding and the weekly working rhythm.
- Earned media covers press relationships, story development, executive access and proof of relevant coverage.
- AI visibility covers prompt research, citation source mapping, answer accuracy and the programs that build authority in sources AI engines retrieve.
- Integrated execution covers how paid, organic, influencer, creator, search and content work is coordinated with PR.
- Crisis readiness covers scenario planning, escalation paths, message discipline, monitoring and response speed.
- Measurement covers outcome dashboards, source-quality analysis, share of voice, citation share and business reporting.
How should you weight the RFP scorecard?
A balanced scorecard gives the largest weight to strategy and the next largest to earned media and AI visibility, then splits the remaining points across execution, crisis readiness and measurement. The weights below are 5W's recommended starting point, and a brand with a regulated product or a recurring crisis risk should raise the crisis weight.
| Capability | Weight | What to score | Evidence to request |
|---|---|---|---|
| Strategy and category intelligence | 25% | Understanding of the market, buyers, competitive set and constraints | A short category diagnosis written for your brand |
| Earned media and narrative development | 20% | Credible, timely, journalist-ready stories with clear proof | Three priority narratives and relevant coverage samples |
| AI visibility and source authority | 20% | Diagnosis of answer-engine visibility, source gaps and hallucination risk | A written methodology and results from your own prompt test |
| Integrated execution | 15% | Earned, social, digital, creator, paid and content work running without silos | A first-quarter work plan showing who owns each channel |
| Crisis and reputation capability | 10% | Speed, calm and accuracy when stakes rise | A crisis process and the results of a tabletop exercise |
| Measurement and operating discipline | 10% | Reporting, meeting cadence, team structure and decision rights | A sample dashboard and a named team structure |
Why do AI answers matter before the RFP is sent?
AI answers matter because they now influence which agencies reach a shortlist before any RFP is issued. An engine such as ChatGPT or Perplexity generates each answer at the moment of the question, drawing on the sources it retrieves, so the agencies it names depend on what those sources say about them.
You can see this yourself. Ask four engines which PR agencies work with brands in your category and record who is named, in what order and with which cited sources. The results will differ between engines and between runs, which is why a single screenshot from an agency pitch is not evidence.
Academic work supports treating visibility as something that varies by domain. Aggarwal and colleagues, in a paper accepted to KDD 2024, reported that optimization methods raised visibility in generative engine responses by up to 40 percent in their benchmark, and that the effect depended on the domain of the query (GEO: Generative Engine Optimization on arXiv). For a buyer, the practical point is to ask each agency for results in your category, not a generic case study.
No agency controls how an engine writes an answer. A credible agency can show how it measures citations, which sources it targets and how it corrects inaccurate descriptions, and it can explain why those actions change what an engine retrieves. 5W's State of AI Search 2026 research and the GEO Reckoning report cover how that retrieval works in more detail.
How do you test a finalist's AI visibility?
You test a finalist's AI visibility by running the same buyer prompts in each engine, logging every answer and repeating the run after two weeks. The steps below take a communications team about half a day and give every agency the same test.
- Write 20 to 30 prompts a buyer in your category would type, covering category questions, comparisons and requests such as "best PR agency for [your category]".
- Run every prompt in ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews on the same day, with chat history and memory turned off where the product allows it.
- Log whether each finalist is named, its position in the answer, how the answer describes it and which sources are cited.
- Repeat the full run after two weeks and note which results held, because generated answers shift between runs.
- Share the log with each finalist and ask it to explain the gaps and what it would change in the first 90 days.
Sample columns for the logging sheet
Use one row per prompt and engine, with columns for prompt text, engine, date, agencies named, position, description used, sources cited and a flag for any inaccurate fact. A separate tab can hold the second-run results so differences are easy to compare. The AI visibility audit glossary entry describes the same method at larger scale.
Which questions should you ask every agency?
Ask every agency the same ten questions in writing before the live session, so answers can be scored against the weights above. The first seven cover the core scorecard and the last three test the AI visibility claims.
- How do you decide which stories are worth pitching and which are better handled through owned, social or direct channels?
- How do you measure brand visibility in ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews?
- Which citation sources currently shape our category, and how would you improve our presence in those sources?
- What would the first 30, 60 and 90 days look like if we started together?
- How do senior leaders stay involved after the pitch?
- How do you handle legal, regulatory, medical, financial or crisis-sensitive review?
- What reporting would we receive, and how would the dashboard change decisions?
- What did our prompt test show about how engines describe us and our competitors?
- How do you correct an inaccurate answer about a client, and how long does a correction usually take to appear?
- Which results can you show in our category, and which parts of your method did not work?
What should a PR agency proposal include?
A strong proposal shows how the agency thinks before it shows what the agency has done. Ask for the following, and set a page limit so the answers stay comparable.
- A brief category diagnosis based on public information about your brand and competitors.
- Three priority narratives with the proof behind each one.
- A first-quarter work plan with named owners.
- A sample measurement dashboard and the reporting schedule.
- The team structure, including who does the daily work and who is senior counsel.
- Relevant case studies with named outcomes, plus the crisis process and AI visibility methodology.
- A clear pricing model. The PR firm cost guide explains the common retainer and project structures.
Leave room for a live strategy session. A conversation shows how the team listens, challenges assumptions and handles constraints, which a written response cannot show.
How do you test crisis readiness?
You test crisis readiness with a short tabletop exercise during the finalist stage. Give each agency the same fictional scenario, such as a product safety complaint going viral on a Friday evening, and ask for a holding statement, an escalation path and a monitoring plan within 24 hours.
Score the speed, the accuracy of the draft and the questions the team asks. Include one AI-specific step: ask how the agency would check what ChatGPT and Perplexity say about the incident on day one and day seven. 5W's crisis communications practice page describes the components of a crisis program.
What are the red flags in agency responses?
The main red flags are guarantees about AI answers, evidence taken from a single prompt and reporting that measures activity instead of outcomes. The table lists common claims and the follow-up question that tests each one.
| Claim in the response | Why it needs a follow-up | Question to ask |
|---|---|---|
| "We guarantee placement in ChatGPT" | No agency controls how an engine generates an answer | How do you measure change, and what happens if it does not move? |
| A screenshot of one prompt as proof | Answers differ between runs and engines | How many prompts, engines and dates does this result cover? |
| Reports built on impressions or ad value equivalents | These count output, not outcome | Which outcome metrics will you report each month? |
| Generic case studies from unrelated categories | Results vary by category and query | Which results do you have in our category? |
| Senior leaders appear only in the pitch | Daily work may go to a junior team | Who attends the weekly call and who writes the first drafts? |
How long should the RFP process take?
A well-run agency RFP takes about six weeks from brief to decision. The timeline below is 5W's suggested schedule, and smaller brands can compress it as long as the prompt test and the live session stay in.
| Week | Step | Output |
|---|---|---|
| 1 | Write the brief and lock the scorecard weights | Approved brief and scorecard |
| 2 | Send the RFP to four to six agencies | Questions log and agency Q&A call |
| 3 | Receive written responses | Scored first pass |
| 4 | Run the AI visibility test and reference calls | Prompt log and reference notes |
| 5 | Hold live strategy sessions with two or three finalists | Final scores |
| 6 | Decide, negotiate scope and announce | Signed scope of work |
How do you measure the agency after hiring?
You measure the agency against outcomes agreed in the scope of work: quality of coverage, share of voice, citation share, accuracy of AI descriptions and business results. Review the dashboard monthly and re-run the prompt test every quarter, since generated answers change faster than search rankings.
Tie each metric to a decision. If AI engines describe the brand with an outdated fact, the agency should name the source that carries it and the action that corrects it. 5W's future of communications measurement report covers the metrics communications leaders are moving toward.
Frequently asked questions
How many agencies should receive a PR RFP?
Four to six agencies is enough for most brands. Fewer than four limits comparison, and more than six adds review time without improving the decision.
Should an RFP include an AI visibility test?
Yes. Running the same prompts for every finalist shows how engines currently describe each agency and gives you a factual basis for questions about methodology and results.
What does a PR agency RFP scorecard weigh most?
Strategy and category intelligence carries the largest weight in this guide at 25 percent, followed by earned media and AI visibility at 20 percent each.
Can an agency guarantee a brand will appear in ChatGPT?
No. Engines generate answers per query from the sources they retrieve, and no agency controls that process. An agency can measure citations, improve the sources an engine retrieves and report the change.
How often should you re-test AI visibility?
Re-test every quarter at minimum and monthly in fast-moving categories, because citation patterns and answers change between runs and model updates.
Where do the scorecard weights come from?
The weights are 5W's recommendation, not an industry standard. 5W is a PR, creative and AI Visibility agency and could be one of the agencies you evaluate, so adjust the weights to your own priorities before sending the RFP.
Related reading
- PR agency AI visibility report
- Generative engine optimization practice
- AI visibility agency overview
- Public relations agency services
About 5W
5W is a PR, creative and AI Visibility agency founded in 2003 and headquartered in Miami, with offices in Tampa and New York. The firm combines earned media, digital marketing, crisis communications, generative engine optimization and proprietary research. For more information, visit www.5wpr.com.
November 2026, 5W Research. Published by 5W. 5wpr.com · research@5wpr.com.
