How Long Should a Headline Be?
The finding
The median press release headline runs between 88 and 98 characters, or 13.0 to 14.8 words, across 134,126 headlines. That is roughly double the length of a typical news headline, and well past the point where search results and email subject lines truncate.
Length by service
| Service | Median characters | Mean words | Over 120 characters |
|---|---|---|---|
| Service A | 98 | 14.7 | 23.7% |
| Service F | 97 | 14.8 | 21.3% |
| Service B | 97 | 14.1 | 20.9% |
| Service C | 93 | 13.2 | 0.2% |
| Service E | 91 | 13.5 | 17.8% |
| Service D | 88 | 13.0 | 24.3% |
The ceilings differ sharply. Service C's longest headline in nearly 44,000 releases was 168 characters, and only 0.2% exceeded 120. Service A allowed up to 460 characters, with 23.7% over 120. That is a platform rule, not an editorial preference — and it means the same headline can be truncated on one service and published whole on another.
What actually survives
Truncation is where headline length stops being theoretical. Search results generally cut around 60 characters. Email subject lines on mobile cut earlier. A 98-character headline is therefore read in full by almost nobody — most audiences see the first half and decide from that.
This is the practical case for front-loading: whatever makes the announcement matter should appear before the cut, not after a company name, a product name and a qualifier. The most common structural mistake is spending the visible half on identity and the invisible half on news.
The two audiences a headline now serves
A press release headline used to have one reader: a journalist scanning a feed. It now has two, and the second one changes the calculus.
The first is still the human on deadline, who reads the truncated version and decides in about a second. That reader is served by front-loading and specificity.
The second is machine: search indexes, monitoring alerts and AI assistants that summarise and cite. These readers do not truncate, but they do extract, and what they extract is whatever is stated most plainly. A headline built around a concrete claim with a number in it is far more likely to be quoted back than one built around a brand name and a superlative.
The useful part is that both audiences want the same thing. Say what happened, attach the number, put it first. There is no trade-off to manage here, which is unusual and worth exploiting.
What to do about it
- Put the news in the first 60 characters. Assume the rest is invisible in most contexts.
- Lead with the thing that happened, not the company that did it.
- Check your service's ceiling before writing to a length it will silently cut.
- Read it truncated. Cut your headline at 60 characters and see whether it still says anything.
The two audiences a headline now serves
A press release headline used to have one reader: a journalist scanning a feed. It now has two, and the second one changes the calculus.
The first is still the human on deadline, who reads the truncated version and decides in about a second. That reader is served by front-loading and specificity.
The second is machine: search indexes, monitoring alerts and AI assistants that summarise and cite. These readers do not truncate, but they do extract, and what they extract is whatever is stated most plainly. A headline built around a concrete claim with a number in it is far more likely to be quoted back than one built around a brand name and a superlative.
The useful part is that both audiences want the same thing. Say what happened, attach the number, put it first. There is no trade-off to manage here, which is unusual and worth exploiting.
Methodology
These findings come from an analysis of 134,126 press releases studied between 7 May and 15 August 2026 across ten distribution services. Body text was available for 129,742 of them, and only services carrying more than 2,000 releases are reported.
Percentages are calculated within each service rather than across the whole corpus, so they are unaffected by how many releases each service contributed. Services are reported anonymously and the same letter refers to the same service throughout; the spread between them is the finding, not the ranking.
Some services display contact details in a masked form rather than plain text; masked details are counted as present. A publication date is not recorded for every release, so analyses that depend on dates or times state their own smaller sample size. Figures last verified 15 August 2026. You are welcome to cite these findings with attribution to PR Gun.
Ready to distribute your press release?
Get your news in front of thousands of journalists with PR Gun's distribution network.