Cite as: EPR Editorial Team. "Do Press Releases Still Work? The Press Release Retrieval Study 2026." Everything-PR Research, August 2026.
Of 250 press releases distributed via PR Newswire and Business Wire in Q1 2026, 47 were cited by at least one AI engine within 90 days. 203 were never retrieved. This is the first study to measure whether wire-distributed press releases actually appear in AI engine answers, what separates the ones that do from the ones that don't, and what to do about it.
Do Press Releases Still Work in 2026?
Yes, for 19% of releases. The wire still works for M&A announcements (40% retrieval), regulatory filings (30%), and product launches with editorial pickup (14%). It does not work for awards, certifications, or minor partnerships: zero of 50 such releases in this study were cited by any AI engine. The gap is not budget, distribution tier, or industry. It is editorial pickup, a citable number in the lede, entity density, regulatory authority, and a schema-marked boilerplate.
Press Release Retrieval Rates by Type
| Release Type | Sample | Retrieved by AI | Rate |
|---|
| M&A / Acquisition | 50 | 20 | 40% |
| Regulatory / SEC Filing | 50 | 15 | 30% |
| Product Launch | 50 | 7 | 14% |
| Executive Appointment | 50 | 5 | 10% |
| Partnership / Award / Other | 50 | 0 | 0% |
How Long Does It Take for a Press Release to Appear in AI?
At 30 days, only 20 of 250 releases had been cited. At 60 days, 35. At 90 days, 47. After 60 days, a release that hasn't been cited is unlikely to be. The retrieval window for a press release is approximately 60 days.
| Window | Cited (of 250) | Cumulative % | New in Window |
|---|
| 30 days | 20 | 8% | 20 |
| 60 days | 35 | 14% | 15 |
| 90 days | 47 | 19% | 12 |
What Makes a Press Release Get Picked Up by AI Engines?
Five predictors separate the 19% that get cited from the 81% that don't.
| Predictor | When Present | When Absent | Multiplier |
|---|
| Editorial pickup by top-50 publication | 38% | 6% | ~6x |
| Citable number in first 100 words | 28% | 9% | ~3x |
| 5+ named entities in body | 27% | 11% | ~2.5x |
| Regulatory / SEC-required | 30% | 15% | ~2x |
| Schema-marked boilerplate | 24% | 13% | ~2x |
Editorial pickup is the dominant variable and it is not close. The engines are not reading the wire. They are reading the journalist who read the wire. The release is the input. The editorial coverage is the citation anchor.
Which AI Engines Cite Press Releases?
| Engine | Cited (of 47) | Cited Wire URL | Cited Editorial |
|---|
| Perplexity | 40 (85%) | 7 | 33 |
| ChatGPT | 28 (60%) | 2 | 26 |
| Gemini | 22 (47%) | 0 | 22 |
| Claude | 17 (36%) | 0 | 17 |
| Google AIO | 14 (30%) | 0 | 14 |
Perplexity is the only engine that reads wire syndication URLs directly. For every other engine, the wire URL was almost never the cited source. Engines cited the journalist who covered the release, not the wire itself.
Which Releases Should Go on the Wire in 2026?
The decision is not whether to write a press release. It is whether a given release earns wire spend, and the retrieval data above sets the threshold.
- Public company material disclosure: wire, every time. Regulation FD requires it.
- M&A, financial restatements, regulatory actions: wire, for the documented timestamp and the 40% and 30% retrieval rates above.
- Executive appointments at the C-suite or board level: wire usually justifies the cost.
- Major product launches with cross-industry implications: wire for syndication breadth, paired with direct pitching since product launches alone retrieve at only 14%.
- Industry partnerships involving named institutional players: wire.
- Survey results, white papers, brand campaigns: corporate newsroom plus targeted outreach, not wire.
- Funding rounds: later-stage rounds typically go wire; seed and Series A increasingly skip it.
- Awards, certifications, minor partnerships: corporate newsroom only. This study found zero of 50 releases in this category cited by any engine.
The cost-benefit calculation has shifted because the AI-retrieval value of wire distribution has risen while the journalist-acquisition value has declined. The wire is documentary and AI-indexing infrastructure first, a journalist-distribution channel second.
How to Write a Press Release That AI Engines Actually Cite
The old press release was written to get one reporter to pick up the phone. The release that earns AI citation is written to be quoted when a prospect asks a model a question. Six rules follow directly from the data above.
1. Lead with a number. "$2.4 billion acquisition" beats "is pleased to announce a strategic transaction." A citable number in the first 100 words carries roughly a 3x multiplier and costs nothing to include.
2. Write the headline as a prompt. "Five Things to Know About the 2026 Launch" answers a query. "Company Announces Strategic Initiative" answers nothing. Write the headline as the answer to the question a buyer would actually type.
3. Name five entities. Companies, people, products, places, regulators. Five or more named entities in the body carry roughly a 2.5x multiplier because models extract and retrieve named entities more reliably than paraphrased claims.
4. Write the first paragraph as a standalone citation. Models pull the lede. It should read correctly quoted on its own, with the who, what, when, and the key number all present, and no "as a leader in the space" filler.
5. Earn editorial pickup. This is the dominant variable in the data, at roughly 6x. The wire delivers the release. The journalist delivers the citation. Pair every wire release with a targeted pitch to five to ten reporters who cover the beat.
6. Build the boilerplate for machines. A schema-marked boilerplate with company name, founding year, headquarters, sector, and CEO carries roughly a 2x multiplier over a generic tagline, and the consistency compounds across every release that reuses it.
The Anatomy of a Press Release That Retrieves
Every section of a press release does a specific job for both the journalist and the model reading it.
Dateline and date. City and date of distribution, for example "PALO ALTO, CA, August 15, 2026."
Headline, 8 to 12 words. News-shaped and searchable. "OpenAI Launches GPT-5 With Multimodal Reasoning for Enterprise AI" states the news. "OpenAI Announces Exciting New Innovation" does not.
First paragraph, 2 to 3 sentences, under 40 words. Answers who, what, when, where, and why. This paragraph works as both the journalist's lede and the model's citation snippet.
Body, 3 to 4 paragraphs. Paragraph two expands on what the news means for customers or the market. Paragraph three carries a quote from a named executive that states a real position, not "we are excited to announce." Paragraph four adds supporting specifics: data, use cases, pricing, availability.
Boilerplate, 50 to 100 words. A standard company description, updated annually and kept consistent release to release so the repetition itself becomes a signal.
Media contact. Name, email, phone.
How Should a Press Release Be Distributed?
A release earns retrieval through layered distribution, not the wire alone.
- Wire service distribution. PR Newswire or Business Wire for every release that clears the bar above. This is what gets the release into search indexes and AI training data at scale.
- Direct media outreach, simultaneous with the wire. A tailored pitch to the specific journalists and analysts who cover the category, often under embargo.
- Owned media, same day. The announcement published on the corporate newsroom, schema-marked and treated as the canonical source URL.
- Social amplification, within 24 hours. A short version on LinkedIn in the executive's voice and on the company's other channels, linking back to the full release.
How Much Does a Press Release Cost vs. Its AI Retrieval Value?
A national wire distribution costs $3,000 to $5,000 per release. At an 81% zero-citation rate, the majority of wire spend produces no AI retrieval value. The cost per citation for the 19% that do get retrieved runs approximately $16,000 to $26,000 per cited release. For partnership and award releases, at 0% retrieval, the AI-retrieval value was zero for every single one: $175,000 to $250,000 in aggregate wire spend with zero return in the AI layer.
This does not make the wire worthless for those releases. Disclosure requirements, newsroom record, and syndication footprint still serve other purposes. But if AI retrieval is the KPI, the reallocation math is clear: wire only the releases that earn pickup, and redirect the rest to newsroom infrastructure and editorial relationships.
How Do You Measure Whether a Press Release Worked?
Four measurements, checked in order.
- Editorial pickup. Which named outlets ran the story, and are any of them the top-50 publications that carry the 6x multiplier.
- Search ranking. Two weeks after distribution, does the release or the editorial coverage of it rank on page one for the release's core terms.
- AI citation, checked at 60 days. Ask ChatGPT, Claude, or Perplexity the question a buyer would ask. Is the release or its coverage the cited source. The retrieval curve in this study flattens after 60 days: a release not cited by then is unlikely to be cited later.
- Referral traffic and brand search lift. Clicks to the site from the release and its coverage, and the change in branded search volume in the two weeks after distribution.
Frequently Asked Questions
Do press releases still work in 2026?
Yes, for 19% of releases. M&A and regulatory releases with editorial pickup achieve 30 to 40% retrieval rates. Awards and partnerships produce zero AI citations. The format remains the most machine-readable corporate communications format, but the wire alone is not enough.
Should companies stop using the wire?
Not for M&A, regulatory, or C-suite moves, which retrieve at 30 to 40%. For partnerships and awards, at 0% retrieval, use a corporate newsroom plus targeted outreach instead.
How do I get a press release cited by ChatGPT?
Lead with a number in the first 100 words, for roughly a 3x multiplier. Include five or more named entities, for roughly 2.5x. Earn editorial pickup from a top-50 publication, for roughly 6x. The engine cites the journalist who covered the release, not the wire URL itself.
Which AI engine cites press releases most?
Perplexity cited 85% of the retrieved releases in this study, the only engine that reads wire syndication URLs directly. ChatGPT cited 60%, but almost always through editorial coverage, not the wire URL.
How much does a press release wire distribution cost?
PR Newswire and Business Wire national distribution runs $3,000 to $5,000 per release. At the 19% retrieval rate documented in this study, the cost per AI citation is $16,000 to $26,000. For more on PR pricing, see How Much Does a PR Firm Cost in 2026.
What replaces wire distribution for a release that skips the wire?
A schema-marked corporate newsroom as the canonical source, owned amplification through LinkedIn and founder channels, targeted direct outreach to the journalists and analysts who cover the category, and pre-briefing of the trade press and newsletters that shape the citation surface for that industry.
Part of EPR's PR industry research series. Related: PR Media Pitching Citation Share Index 2026 · The Journalist AI Adoption Study 2026 · How to Write a Press Release That Gets Coverage · The Japanese Press Release Wire That's Now a Citation Anchor
Methodology: 250 press releases, stratified random sample, PR Newswire and Business Wire public feeds, Q1 2026. 50 per release type, 50 per industry. Each tested against five engines at 30, 60, and 90 days. 15,000 individual observations. Dual-rater scoring, 94% inter-rater agreement. Full dataset available on request.
