The standard PR agency pitch has not changed much in twenty years. You sign a retainer, pay every month whether placements land or not, and receive activity reports that describe effort rather than outcomes. The agency builds institutional knowledge. You build a habit of paying for potential.
In 2025, 90% of digital PR agencies were still using the monthly retainer as their primary billing model, according to a BuzzStream survey of digital PR practitioners. That figure is not changing because the retainer is working. It is not changing because the retainer is the default, and defaults are hard to break.
Something else changed instead. The reader did.
AI systems now conduct the first round of brand research for most B2B buyers. When a founder or CMO asks Perplexity or ChatGPT who the best PR agency is, or which cybersecurity company leads its category, the answer comes from a synthesis of sources the model has indexed and trusts. Those sources are not brand websites or paid ads. According to the Fullintel-UConn academic study published at the International Public Relations Research Conference in February 2026, 89% of links cited in AI responses were earned media. Ninety-five percent were unpaid.
That single shift makes the pay per placement model the most structurally sound option in the PR market today. Not because it feels fairer to clients. Because it is the only model where the agency's incentive and the client's actual outcome are pointed at the same thing.
The mechanics are simple. A pay per placement PR agency works to secure editorial coverage in publications. The client pays only when a placement goes live. No placement, no invoice. The agency absorbs the risk of failed outreach, unresponsive editors, and news cycles that bury the pitch.
This is the opposite of the retainer structure. A retainer transfers risk to the client immediately. You pay from day one regardless of what publishes. The agency captures revenue whether the month produces placements or a series of "we're working on it" updates. According to the PR Council billing rate research, the average hourly billing rate across all PR titles reached $278 in 2025, up 7% from 2023. Those hours run whether or not a single article appears.
The pay per placement model also creates a different selection dynamic. When an agency's revenue depends on placements clearing, they are selective about which clients they take on and which stories they bring to editors. Agencies operating on retainer have less structural pressure to be selective; their revenue does not depend on win rate.
This is not a moral claim about retainer agencies. It is a mechanical observation about incentives. The pay per placement model forces quality discipline because the agency's economics require it.
For most of PR's history, the case for earned media rested on human reach: impressions, readership, coverage volume, brand awareness. Those metrics were always approximate and always contested. The connection between a Forbes article and a closed deal was causal but murky.
AI search changed the causal chain.
When Bain's 2025 research on AI search behavior found that roughly 80% of search users now rely on AI summaries at least 40% of the time, and about 60% of searches end without the user clicking through to a source, the mechanism became clear. AI systems are not just surfacing links. They are replacing the click. A brand that appears in the AI summary gets the recommendation. A brand that does not appear gets nothing, not even the option of a good headline driving traffic.
The sources AI engines synthesize into those summaries are the same publications PR has always targeted. Moz's 2026 analysis of 40,000 queries in Google's AI Mode found that 88% of AI citations were not in the organic top 10 of traditional search. Forbes, TechCrunch, Reuters, and the Financial Times do not top traditional search results for most queries. But they are the sources AI engines treat as authoritative. The Muck Rack Generative Pulse report from December 2025 found that 82% of links cited by AI engines were earned media, with Reuters, the Financial Times, Forbes, Axios, and Time among the top cited outlets.
This is the new math. An article in Forbes is not just a brand awareness moment. It is a training input for every AI engine that indexes it. It is a citation source for every AI response that touches your category. It compounds over time as models update and refine their understanding of which sources speak authoritatively about which topics.
Pay per placement agencies, by definition, deliver the input. Retainer agencies may or may not deliver the input, and they charge either way.
The argument for retainers has always been that reputation takes time and that agencies need continuity to build relationships and execute long-term narratives. That argument has merit. The problem is that it has been used to justify a billing model that disconnects payment from delivery.
| Model | When client pays | Risk held by | Agency incentive | AI visibility outcome |
|---|---|---|---|---|
| Monthly retainer | Every month regardless of placements | Client | Maintain the relationship | Variable; depends on execution discipline |
| Pay per placement | Only when placements publish | Agency | Secure placements at highest-authority publications | Direct: every invoice = one AI citation source |
| Project-based | At project completion | Shared | Complete the defined deliverable | Inconsistent; depends on what the project is |
| Hybrid (retainer + performance bonus) | Monthly base + placement bonuses | Mostly client | Exceed defined milestones | Better than pure retainer, weaker than pay per placement |
The Agency Reporter's 2026 analysis of PR pricing models put it plainly: clients are not rejecting PR, they are rejecting vague value. The retainer's defense (that PR work happens between the visible moments, in strategic counsel and issue anticipation) holds in large enterprise relationships. It does not hold for founders and growth-stage CEOs who are paying $5,000 to $15,000 per month and asking what, specifically, that money produced last quarter.
The Forrester State of Business Buying 2024 found that 70% of B2B buyers complete the majority of their research before their first contact with a vendor. If that research happens in AI systems, and AI systems build their answers from earned media placements, then every month a company pays a retainer without securing placements is a month where their AI visibility did not improve. The retainer model offers no structural protection against that outcome. The pay per placement model does.
Not all pay per placement offerings are equivalent. The term describes a billing structure, not a quality tier. A placement in a regional business journal and a placement in the Wall Street Journal are both "placements." They are not comparable inputs for AI visibility purposes.
The relevant variable is publication authority. Ahrefs' analysis of ChatGPT's citation behavior found that 65.3% of cited pages came from DR80+ domains, high-authority sites with established trust signals. A pay per placement agency that places clients in DA-40 aggregators is delivering a billing structure change, not an AI visibility improvement.
The publications that consistently appear in AI engine responses for B2B categories are the same ones that have driven PR value for decades: Forbes, TechCrunch, the Wall Street Journal, Fast Company, Inc, Wired, Business Insider, the Financial Times, Reuters. Securing placements in these publications requires direct editorial relationships, not software, not press release distribution, not outreach automation.
The Yext research published in January 2026, analyzing 17.2 million distinct AI citations across ChatGPT, Gemini, Perplexity, Claude, SearchGPT, and Google AI Mode, found distinct citation patterns by model. Gemini favors first-party brand sites. Claude cites user-generated content at 2-4x higher rates than other models. No single optimization strategy works across all platforms.
What does work across all platforms is what it has always been: actual published articles in publications AI engines index and trust. The model that guarantees delivery of that input, pay per placement with editorial relationship-based execution, is the only structural answer to a multi-platform AI citation environment.
The billing structure is necessary but not sufficient. A pay per placement agency that places clients in low-authority syndication networks has given you the model without the outcome. Evaluate on these factors:
Every Tier 1 placement is not just a coverage moment. It is a node in the AI citation graph. When an article about your company's approach to enterprise security appears in TechCrunch, every AI engine that has indexed TechCrunch now has a data point linking your company to that category. When a prospect asks ChatGPT who the credible enterprise security companies are, the answer is built from those nodes.
This is what the discipline of Machine Relations describes at the infrastructure level: earned media placements in publications AI engines trust are not a marketing tactic. They are the mechanism by which AI systems decide which brands to surface, recommend, and cite.
The AT research on earned versus owned AI citation rates found a 325% differential in AI citations between brands with consistent earned media presence and those relying on owned distribution. That gap is not closed by a SaaS tool, a schema optimization, or a structured data audit. It is closed by placements in publications AI engines treat as authoritative sources.
Pay per placement is the billing model. Earned media at scale in the right publications is the mechanism. The two are aligned because you cannot invoice without the placement, and you cannot build AI citation authority without the placement. The retainer model breaks that alignment by making payment independent of delivery.
This is also why the retainer model's defense (that relationship-building and strategic counsel have value between visible moments) is less persuasive in 2026 than it was in 2016. When Gartner projected in 2024 that traditional search volume would decline 25% by 2026 due to AI chatbots, and Pew Research found that click-through rates halved when AI summaries appeared, the invisible PR work between placements became less defensible as a line item. What moves the number is what gets published. A billing model that ties payment to publication is built for that reality.
One distinction that gets obscured in the AI era is the difference between agencies that build editorial relationships and agencies that build software on top of press release databases.
Press releases have their uses. The Muck Rack Generative Pulse December 2025 found that press release visibility in AI citations grew 5x year over year. But press releases still account for only 1% of AI citations. Ninety-nine percent of AI citation share belongs to actual editorial coverage. An agency whose primary distribution mechanism is press release syndication is competing for 1% of the citation market while offering you the pay per placement billing model on top of it.
Editorial relationships are the relevant asset. Direct access to editors and journalists at Forbes, TechCrunch, the Wall Street Journal, Fast Company, and comparable publications means the agency can pitch a story and have it considered by someone who knows their name. That is categorically different from automated outreach to a media database. The editor relationship determines whether the pitch gets read. The publication's AI indexing determines whether the placement produces citation value. Both matter.
When evaluating a pay per placement agency, the question that cuts through most positioning is simple: who do you know at the publications I care about, and when did you last place a story with them?
Pay per placement PR means the agency charges only when a media placement is secured and published. A retainer charges a fixed monthly fee regardless of whether any placements occur. The structural difference is risk allocation: with pay per placement, the agency absorbs the risk of failed outreach. With a retainer, the client pays whether or not placements materialize.
Yes, because AI search visibility is built from editorial placements in trusted publications. The Fullintel-UConn IPRRC study found 89% of AI citations come from earned media. A pay per placement model is the only billing structure that guarantees delivery of the input that builds AI citation authority. Retainer agencies may deliver placements, but they charge regardless of whether they do.
Publications with DR80+ domain authority that AI engines consistently index for your category. For most B2B categories, this means Forbes, TechCrunch, Business Insider, Fast Company, Inc, Wired, the Wall Street Journal, the Financial Times, and CNBC. The Ahrefs ChatGPT citation analysis found 65.3% of cited pages come from DR80+ domains. Volume of placements in lower-authority publications does not compensate for absence from high-authority ones.
Ask directly for their placement rate within 90-day engagement windows and ask for a representative sample of client placements from the last quarter. Legitimate agencies with genuine editorial relationships can produce this information. If the answer is vague or framed around "campaigns" rather than specific publication placements, treat that as a signal about their model's actual performance.
Both. Pay per placement works for launches because it aligns payment with deliverables. It works for long-term brand building because AI citation authority compounds over time: each placement in a Tier 1 publication is a permanent node in the citation graph. The compounding effect is actually better served by pay per placement than retainer because it sustains production discipline across months, not just campaign windows.
The pay per placement conversation in PR has historically been framed as a client-friendly innovation, a way for brands to reduce risk and improve accountability. That framing is accurate but incomplete.
In 2026, the more important fact is that pay per placement is structurally aligned with how AI systems build their knowledge of which brands are credible, which companies lead their categories, and which sources can be trusted to synthesize into a recommendation.
PR's original mechanism (earn coverage in publications that your buyers trust, and let that third-party credibility do the conversion work) was always sound. The publications that shaped human brand perception for decades are the same publications AI engines treat as authoritative sources. What changed is that the first reader is now often a machine, and the machine does not click through to the brand's website. It cites the article. The brand that has more articles in more trusted publications is the brand the machine recommends.
Machine Relations is the name for this shift: the discipline of ensuring your brand is consistently surfaced, cited, and recommended by AI systems through earned media in publications they trust. The pay per placement model is the billing structure most aligned with that discipline because it ensures the agency's revenue depends on the same thing the client's AI visibility depends on: actual published placements in actual publications.
If you want to see where your brand currently stands in AI search answers and which placements would move that number fastest, start with a visibility audit.