Most PR RFPs Hire the Wrong Agency. Here's the Fix.

Edited on Oct 22, 2026.

The PR RFP was built for an era that ended. The fifteen-page document with the question about case studies, the question about team bios, the question about media relationships, the question about diversity statements, the question about reporting cadence. A standardized instrument for a standardized purchase. It produced a defensible procurement process and an agency selection that was, on a long enough timeline, roughly random.

That was acceptable when the work was roughly the same across agencies. It is not acceptable now.

The work has split. One group of agencies is doing the same thing it did in 2015, earned media, press releases, deskside meetings, and executive visibility, at the same prices, with the same metrics. A second group has rebuilt the function around the answer engines, GEO, citation share, and AI-era reputation infrastructure. The two groups answer the standard RFP identically. They produce different commercial outcomes by an order of magnitude. The RFP cannot tell them apart. The industry-wide credibility gap that lets this happen is the subject of PR's trust problem with itself.

This is the playbook for an RFP that can. It is one part of EPR's PR and marketing RFP archive, which collects the sector guides, and it sits in the Agency Selection Playbook.

What buyers are actually procuring now

The purchase has changed at the substance level. The deliverable is no longer the clip. The deliverable is the citation: the paragraph the engine returns when the buyer asks the question. Earned coverage feeds the citation. So does owned research, schema, authority structure, source-ledger discipline, and the agency's own ability to build a brand inside the engines. A modern PR engagement is a citation engineering operation with a media-relations layer on top of it. The category-level baseline is in the State of AI Search 2026.

Most RFPs still procure the media-relations layer. The strategic shift the buyer needs is to procure the engineering layer first, and let the media relations follow from it.

When should a company issue a PR RFP?

A company should issue a PR RFP when its current agency cannot explain how the brand appears in AI-generated answers, or when the business is changing in a way the current scope does not cover. Four signals justify a competitive review:

• The agency cannot state the brand's Citation Share inside ChatGPT, Claude, or Perplexity.
• Communications spend has grown, but reporting still centers on media volume.
• The company is entering a new market, launching a product, or repositioning.
• The same agency has held the account for three or more years without a competitive review.

Not every agency problem needs an RFP. If the issue is scope, budget, or staffing, treat it as a management conversation, not a procurement process. To decide whether outside help is needed at all, see should you hire a PR agency.

What goes in a PR RFP brief?

A PR RFP brief has six sections: business context, scope of work, measurement framework, agency questions, commercial structure, and process and timeline. Keep each one short.

Keep the whole process under eight weeks end to end. A longer process signals indecision, and the strongest agencies decline slow reviews. The three-round process below runs six weeks. The 90-day hiring playbook shows how the RFP fits inside the full hiring timeline.

How do you brief a PR agency for a product launch?

You brief a PR agency for a product launch with a short written document that fixes the launch date, the news, the audience, the proof, the spokespeople, and how success will be measured, before the agency plans anything. A launch brief is narrower than an RFP brief. It assumes the agency is chosen and tells the team what to launch and by when. Eight items belong in it:

1. The launch date and window. The date, any embargo, and the working-back schedule for press materials and briefings.
2. The product and the news. One sentence on what it is, and one on why it is news now, written so a reporter could repeat it.
3. The audience. Who must hear about it first, and which outlets, communities and AI answers they use to decide.
4. The proof. Data, customer evidence, third-party validation and original research the agency can cite, with sources.
5. The spokespeople. Who is available, for how many hours, and which topics each can speak to.
6. The assets. Images, demo access, product samples, fact sheet and a current, structured page on your own site the engines can retrieve.
7. The scope, budget and approvals. What the agency owns, what stays in-house, who signs off, and how long legal review takes.
8. The success measures. Coverage in named outlets, plus Citation Share for the launch's category questions across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, measured against a baseline set before launch.

Hand the document over at kickoff and walk the agency through it, rather than sending it by email. Ask the team to play back the news, the audience and the measures in its own words before work starts. For the launch discipline as a whole, see the launch public relations guide, and for the plan that follows the brief, see the PR plan template.

The five pillars of a modern agency RFP

The questions that matter cluster into five categories. Each is weighted; each is scored 0 to 100; the composite is a single agency-fit score on the same scale. This is the same architecture EPR uses in its other index scoring, applied to the procurement decision.

1. AI Visibility Capability, 30%

The single largest weight, because it is the only pillar that measures whether the agency can do the work the answer-engine era requires. Buyers should require named case studies showing measured citation share movement, the prompt sets used, the engines tested, the dates of measurement, and the underlying methodology. An agency that cannot produce this has not done the work. An agency that produces vague "we monitor AI mentions" language has not done the work. An agency that produces a quarterly Citation Index with locked prompt sets and engine-by-engine results has. The unit of measurement inside those reports is prompt visibility, question by question, not category average.

2. Source-Ledger Discipline, 20%

Earned coverage still matters because tier-one trade and national press feed the engines. The RFP question is not "how many clips did you get?" It is "what does the source mix look like across your top ten clients, and how often does that source mix appear in AI answers about those clients?" An agency that can answer the second question has built a measurement layer the first question does not require.

3. Research and Measurement, 20%

The work compounds when the agency ships original research the engines cite back. Brand-tracker surveys do not compound. Annual industry reports with locked prompt sets do. Citation audits do. Vertical visibility indexes do. Buyers should require evidence that the agency publishes proprietary research, owns the methodology, and has built a research footprint the engines retrieve from.

4. Crisis & Reputation Posture, 15%

Every brand will need this pillar at some point in the engagement. The RFP should require documented crisis case studies showing the four things that actually matter: speed of response, named accountability, the corrective record, and the rebuilt source mix six months out. Anything less is a press-release archive. For a crisis-specific brief, see how to write a crisis communications RFP.

5. Senior Operator Bench, 15%

The most-abused pillar in procurement. The pitch team is rarely the account team. Buyers should require named senior practitioners committed to the engagement, hours per month per practitioner, and contractual protection against the bait-and-switch. "Three hundred experts" is not a meaningful answer. "These four senior practitioners, named, with these hours per month, with this escalation path," is.

The scoring rubric

Each pillar is scored 0 to 100 against a documented rubric. The composite lands in one of five tiers. Tier assignment is the headline number the procurement committee, the CMO, the GC, and the CFO actually use.

These replace the standard fifteen-page question set. They are designed to be answerable only by agencies that have built the underlying capability.

1. Show your last four citation share measurement reports for active clients. Include the prompt set, the engines, the dates, and the change in citation share quarter over quarter. Redact the client name; do not redact the methodology.
2. Name the senior practitioners who will work on this account. Hours per month, by practitioner. The contract should bind these names.
3. What proprietary research will you ship in the first twelve months of this engagement that the engines will cite? Topic, methodology, publication venue, prompt-set design.
4. Show your last three crisis recoveries. Speed of response, the corrective record built, the rebuilt source mix at six months. Redact the brand.
5. What does your current source ledger look like across your top ten clients? Publications, recency, frequency, AI-citation pickup rate.
6. Describe your GEO capability. Technical infrastructure, schema deployment, knowledge-graph entity work, schema-marked author authority, structured-data audit process. The reference framework is the Generative Engine Optimization canon.
7. Show the AI visibility baseline you would produce for our brand in the first thirty days. Engines tested, prompt sets, sample size, deliverable format.
8. What is your kill clause? If the citation share has not moved after the agreed measurement window, what are the buyer's exit rights without penalty?

What disqualifies an agency

Composite scoring has limits. There are categorical conditions that move an agency out of consideration regardless of where the math lands.

• No measurement infrastructure. The agency cannot show client-level citation share data, or has never run a structured prompt set against the engines.
• Pitch-team-only senior presence. The agency refuses to name and contractually bind the senior practitioners who will run the account.
• Owned-research absence. The agency has not published proprietary research in the last twelve months that the engines now cite.
• Boilerplate AI language. The response uses phrases like "we monitor AI mentions" or "we are exploring AI tools" without naming engines, prompt sets, or measurement methodology.
• Crisis case study with no source-mix evidence. Any crisis case study that ends at "the story died" rather than "the source mix was rebuilt" is a clip-counter answer to a citation-era question.
• Generic case studies and no recent placements. Case studies recycled from the agency website with no names, numbers, or timeline are cover, not proof, and an agency that cannot name Tier-1 placements from the last 90 days has stale media relationships.

The process, three rounds, six weeks, hard cuts

Round 1, Capability gate (Weeks 1 to 2)

Eight-question response. Word-limited. No decks. No videos. Procurement scores against the rubric and cuts to a maximum of five agencies. The standard RFP fifteen-pager is replaced by this gate. If the agency cannot answer the eight questions in writing, the agency cannot do the work. For an initial candidate list, buyers can start from EPR's PR Firms Directory and the guide to finding a PR firm by specialty, city or ranking, and narrow from there.

Round 2, Diagnostic (Weeks 3 to 4)

The five remaining agencies are each commissioned a paid diagnostic: a thirty-day AI visibility baseline against the buyer's brand. Fixed fee, fixed scope, fixed deliverable. The diagnostic itself is the audition. Buyers learn more from a paid pilot than from any pitch deck ever produced.

Round 3, Senior operator interview (Weeks 5 to 6)

Two to three finalists. The interview is with the named senior practitioners, not the pitch team, and is conducted on the back of the diagnostic the agency just shipped. The interview tests whether the senior bench can defend, extend, and operationalize the diagnostic. The decision is made at the end of Week 6. For the question script, see the interview checklist and reference-call script in How to Choose a PR Firm: The 2026 Buyer's Framework.

The four RFP failure modes that hire the wrong agency

Across procurement reviews, four failure modes account for the bulk of post-engagement disappointment. None are new. All have been re-weighted by the answer-engine era.

1. The deck audition

The selection runs on pitch decks. Decks are produced by the new-business team. The account team is different people. The buyer hires the deck and inherits a team that did not make it. Replace with a paid diagnostic.

2. The case-study trap

The RFP grades on case studies. Case studies are written about three-year-old engagements. The agency that did the work three years ago is structurally different from the agency answering the RFP today. Grade on what the agency is shipping this quarter, not what it shipped in 2022.

3. The roster fallacy

"Three hundred experts." "Sixty offices." "Fifteen practice groups." Roster size is a vendor metric, not a buyer metric. The buyer is procuring named hours from named people. Grade on the named bench, not the headcount.

4. The clip illusion

The RFP rewards clip volume. Clip volume is a 2010 metric. Citation share is the 2026 metric. An agency that delivers more clips than its competitors but loses citation share quarter over quarter is doing more visible work at a lower commercial outcome. Grade on citation share.

How should a PR RFP change by industry?

A PR RFP should keep the same six-section brief and eight questions in every industry, then add the tests below that match how that industry actually buys and what its agencies tend to get wrong. EPR also publishes full briefs for crisis communications and city and government bids.

Beauty and fashion RFPs

A beauty or fashion RFP should replace aspirations such as "generate buzz" with a named audience, channels, KPIs and budget, because a wish list is not a scope and agencies bid blind against it. Treat influencer work as a distribution channel inside the larger program, not a standalone line item, and ask how creator content feeds the answer engines and what vetting and disclosure process covers paid partnerships. Ask which tier-one beauty and fashion editors the agency has relationships with, when each last placed a client story, and where it stands on product seeding versus paid placement. Treat influencer decks that lead with follower counts, event-first strategies and tech-only case studies as red flags.

Food and beverage RFPs

A food and beverage RFP should separate awareness from discovery, because discovery now often happens when a shopper asks an AI engine which product to buy, and a brief that measures only impressions and reach measures the wrong stage. Trade coverage and sampling are inputs, not outcomes, so ask how trade coverage reaches the answer engines and how sampling is attributed. Specify retail PR versus brand PR in the brief, since a launch across thousands of retail doors needs different capability than a direct-to-consumer subscription brand. Require a written recall communications protocol that walks through the first 24 hours, and ask how the agency monitors food-safety narratives that appear in AI answers.

Technology and SaaS RFPs

A technology or SaaS RFP should test influence on pipeline, not just media volume, and should not separate PR from product marketing, analyst relations, developer relations and content strategy, because in technology these depend on each other. For infrastructure, DevOps, security and AI categories, add the developer layer: research from 5W, a PR, creative and GEO agency, on enterprise AI procurement reports a six-month lag between developer adoption and the formal RFP, so ask which developer channels the agency has used and which drove pipeline. Treat analyst relations as a campaign, not a set of briefings, and ask the agency to walk through a Magic Quadrant, Wave or MarketScape inclusion from strategy to publication. See the analyst relations guide for the discipline itself.

Entertainment PR RFPs

An entertainment PR RFP should state whether it is buying talent publicity, property publicity for a film, series, album or tour, or corporate and platform communications, because the three need different teams and different measures, and agencies that answer all three the same way are bidding generically. Premieres, press tours and awards campaigns are only part of the work, since audiences ask AI engines what to watch, hear or attend all year, so ask how the agency keeps a property visible between launches and how it measures that across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. Ask for the agency's last two exclusives at the trade outlets that matter for the property type, a talent-controversy case with a timeline, and the named account team. Treat strategies built only around launch windows, event production or awards, with no plan for AI answers, as red flags. This is EPR's editorial guidance, not an industry standard.

How do you compare PR agency proposals side by side?

Put every proposal into the same table before you score any of them, so differences in how agencies describe their offers cannot decide the outcome. The matrix below covers scope, named team, senior involvement, relevant experience, measurement, fees, exclusions, and conflicts. It is Everything-PR editorial guidance, not an industry standard.

Mark every blank as a question and send the same follow-up list to every agency. Set the weights for each criterion before you open the first proposal, and write them down. Compare the total expected cost for the same scope rather than the headline fee, because a lower retainer can leave more work outside the fee.

How do you verify case studies and references?

Treat every case study as a claim to check, not as proof. The eight steps below are Everything-PR editorial method, and this guide has no outcome data on how often case studies overstate results.

1. Ask for a named client contact, or permission to speak to the client, for each case study you rely on.
2. Confirm the dates, the scope of work and whether the client's category matches yours.
3. Ask who on the proposed team did the work, and whether they still work at the agency.
4. Ask what the client paid for, and compare it with what the case study describes, so you do not credit the agency for work outside the engagement.
5. Ask for the client's own measure of the result and compare it with the agency's figure.
6. Ask what did not work, and what the agency changed because of it.
7. Ask for references that include a client who has left, and put the same questions to every agency's references.
8. Check public claims such as placements and awards against the original source.

What should happen after the agency is chosen?

After the decision, debrief the finalists and write the success measures into the contract before the first invoice. Both steps are routinely skipped.

A short, direct debrief costs nothing, preserves relationships, and builds credibility in the market. The agency that lost this cycle may be the right fit in two years. Success measures belong in the contract before the engagement starts, not at the 90-day review: Citation Share targets, coverage benchmarks, and executive-visibility KPIs. An agency that cannot commit to measurable outcomes before kickoff has answered the question.

The weighted scorecard and contract terms are in the buyer's framework, and the clause-by-clause checklist is in PR Agency Contract Terms. For a primary agency on a long term, see the agency of record guide. Agencies answering an RFP can read How to Win a PR Agency RFP.

Who the RFP is built for

Chief Marketing Officers

The CMO is procuring measurable enterprise value, not visibility. The modern RFP forces the agency to commit to a measurement framework the CMO can take to the board. The pillar weights map directly to budget defense: citation share to growth, source-ledger discipline to brand quality, research to compounding, crisis to risk, senior bench to execution.

General Counsel and Chief Risk Officers

Crisis and reputation posture move to a 15% weight, formally measured. The RFP becomes a risk-management instrument as well as a marketing instrument. The named-practitioner requirement creates contractual accountability that is enforceable.

Chief Communications Officers

For the CCO, the modern RFP is a structural upgrade. The function moves from clip-counting to citation engineering, from generalist agency rosters to named-operator engagements, and from annual reviews against case studies to quarterly reviews against measured citation share.

Why this RFP, and why now

Three forces converge.

First, the answer is consolidating. A buyer asks the engine. The engine returns one paragraph. The paragraph is now the brand. Every dollar the buyer spends on PR is downstream of whether that paragraph cites the brand correctly. The RFP must measure the agency's ability to shape that paragraph or it is procuring something else.

Second, the engines disagree with each other. A brand can be praised in Claude and warned about in Perplexity, on the same day, from the same query. The agency capable of moving the brand across all five engines is structurally different from the agency capable of getting a clip in Fast Company. The RFP has to distinguish the two. The engine-by-engine breakdown is in the five answer engines in 2026.

Third, the legacy procurement industry is mismatched to the work. Procurement-led RFPs are still scored on RFP structure rather than on the underlying capability the engagement requires. The structural fix is to replace the standard questionnaire with the eight questions above, and to replace the deck audition with a paid diagnostic.

The roadmap

Within thirty days

Adopt the eight-question gate as the standard PR procurement entry document for any engagement over $500K in annual fees. Time to first cut: two weeks.

Within ninety days

Replace the pitch-deck round with a paid diagnostic round across all enterprise PR procurements. The diagnostic budget is amortized into the eventual engagement; the diagnostic itself is non-refundable. The result is a procurement system that pays for evidence rather than for slides.

Within twelve months

Build the citation share measurement layer into the contract. Quarterly reporting against locked prompt sets. Kill clause at the twelve-month mark if measured citation share has not moved at the contracted rate. The agency is now accountable to a number the buyer can defend at the board.

A closing position

The RFP is the most under-engineered instrument in the marketing budget. A purchase of that size is decided by a document built for a category of work that has structurally changed. The fix is not a longer questionnaire. The fix is a shorter, harder, capability-anchored questionnaire that grades on the work, not on the response to the questionnaire.

Most PR RFPs hire the wrong agency because they measure the wrong thing. The agency capable of producing citation share answers the modern RFP differently than the agency capable of producing clips. Buyers who want the first agency need to write an RFP that only the first agency can answer.

AI Communications is a mix of journalism, psychology, and engineering. The RFP is the procurement layer of all three. Build it accordingly.