Every week somebody asks a search engine or an AI assistant some version of "do press releases actually get media coverage?" I run a press release platform, so you would expect me to say yes. I am going to say something more useful: I track what happens to every release we publish, and the honest answer is that almost nobody from the media reads them, and that this matters much less than it used to, because the reader has changed.
I came to this from Shopify, not PR
I did not come to press releases from public relations. I came from Shopify. I spend my time helping shop owners get a foothold in their niche, and the recurring problem is that the established brands own the search results and the ad budgets, so a new store is competing on terrain somebody else already won.
While investigating what could actually move the needle for those clients, I noticed that AI assistants were starting to answer buying questions directly, and that they quote sources rather than rank them. A press release written for AI agents, rather than for journalists, looked like a way for a small business to become one of those sources before the established brands bothered to. That is a leapfrog, not a catch-up. I built Pressonify to test that idea with real tracking rather than take anyone's word for it, including my own.
What the first week of a press release really looks like
The first week of a release has a shape, and it repeats.
Hours one to twenty-four: the crawlers arrive
Within hours of publishing, the AI crawlers turn up. On the day the GPSBob release went live, PerplexityBot fetched the page five times, ClaudeBot twice and Bytespider twice. The Edinburgh Boiler Company release got PerplexityBot three times and OpenAI's search crawler twice on day one. Amazonbot and Meta's agent follow over the next few days. Nobody tells these crawlers a release exists. They find the newsroom and come.
Days two to seven: the quiet gap, then the citations
Being crawled is not being quoted. We check each release against twenty questions a customer might ask an AI engine, on days one, two, seven, thirty, sixty and ninety. The day-one and day-two checks mostly come back empty. The day-seven check is where the citations land.
Of nine releases we published between late September and mid October 2026, eight had citations by the day-seven check, most of them at position one in the answer. The ninth, for PlanningLeads.ie, was first cited on day thirty. Here is the batch, with the citation count at day seven:
- NILA-AV, published 29 September: 13 citations
- Ezi Home Services, 30 September: 8 citations
- Vulnverse, 5 October: 17 citations
- GPSBob, 13 October: 4 citations
- The Edinburgh Boiler Company, 13 October: 4 citations
- PlantGift, 13 October: 3 citations
- The SEO Works, 22 September: 3 citations
- Pressonify.ai, 25 September: 5 citations
Almost all of that is Perplexity. Of the 71 citations we recorded in the last month, 64 came from Perplexity, 5 from Gemini and 2 from Claude.
One caveat I want to keep in: the "day seven" pattern is partly our own schedule. We scan on day seven, so that is when we see it. The honest claim is "within a week", not "on day seven".
And the journalists?
In thirty days our newsroom pages were fetched 3,659 times by AI bots and viewed 412 times by humans. On the three October releases, all but one of the human visits came from a link inside our own site. Not one visit came from a news outlet, and no customer has ever told me a reporter called because of a release.
The closest thing to a human reader in the data is a single visit to the Edinburgh Boiler release on 19 October from ChatGPT-User, which is the fetch that fires when a real person asks ChatGPT a question and it goes and reads the page to answer. One person, somewhere, asked ChatGPT something and got that release as part of the answer. That is what coverage looks like now.
Why this is a leapfrog for small businesses
It works for small businesses because the cost of entry is close to nothing and the incumbents have not shown up yet. A release on Pressonify costs €49.95, and the first one €9.95. The wires charge between $350 and $1,200 for a single release, so for the price of one wire distribution a small shop can publish a year of releases aimed at AI engines. You can see the full comparison in our press release distribution pricing guide.
PlantGift versus Interflora
The "challenge the big players" part is PlantGift versus Interflora. In May 2026 I typed "free plant gift delivery Europe" into Perplexity on a clean device. It recommended PlantGift, citing the shop's own delivery page and the April press release as its two sources, and in the same answer it set Interflora aside as a qualified option and described Bloombox Club as not fully free by default. I reproduced the result on a second device. The full write-up, with the answer text, is in Perplexity recommends PlantGift.
Interflora has decades of brand, a marketing budget I could not match in a lifetime, and every ranking signal Google cares about. None of that mattered to the question being asked, because the answer engine wanted a source that stated the fact plainly, and the small company had published one.
Thin footprint is an advantage
What the scans show is that a company with almost no web footprint often gets cited faster than a well-known one. When an established brand publishes a release, the AI engine already has the brand's own website, its Wikipedia entry and years of coverage to draw from, so the release is one more source competing with the brand's own pages. When a small company publishes, the release can be the only well-structured page on the internet that answers "who is this company and what do they do". There is nothing to compete with, so it gets quoted.
Vulnverse, a company almost nobody had heard of, took 17 citations in its first week. The SEO Works, an established agency with years of presence, took 3. The engine had nothing else to say about Vulnverse, so it said what the release said.
What separates a release that gets quoted from one that does not
The release that gets quoted is the one that answers the question an AI engine is about to be asked, in words it can lift straight out. My rules of thumb, in the order I learned them the hard way:
Put the company name in the headline
Seventeen of the twenty questions we scan are brand questions: "who is X", "X latest news", "X announcement". In October two good releases, GPSBob's EV tracker piece and the Edinburgh Boiler Company's fuel poverty piece, went out with headlines that did not contain the company name at all. After six days they had zero citations, while our newsroom index page, which lists the company name beside each release, was being cited instead. We changed the rule so every headline names the company and the page title keeps it. At the next scan both releases had citations. I would not claim the fix alone did it, but I will not publish a brand-less headline again.
State one fact plainly enough to be copied
The PlantGift fact that travelled into Perplexity's answer was free delivery across Ireland and twenty-five EU countries. A number, a place, a date. Vague claims about being a leader get summarised away. Specific facts get quoted, and the quote is the citation.
Write the About paragraph as the answer to "who is this company"
When an engine is asked who a company is, it looks for a short, declarative description. We now build the release's "Who is X?" answer from the customer's own boilerplate and nothing else. If the customer has not written one, the question is left unanswered rather than invented.
A distinctive name helps more than a big one
The engine had nothing else to say about Vulnverse, so it said what the release said. A company whose name is also a common word, or shared with a bigger business, is fighting for the entity before it can fight for the citation.
Say the same thing on your own site
The PlantGift recommendation came from two sources stacked together: the release and the shop's own delivery page, each confirming the other. A release alone is one source. A release plus an owned page that agrees with it is a fact.
Publish where the crawlers already go
The newsroom took around three and a half thousand AI bot fetches last month. A release there is found within hours. The same text on a company blog that nobody crawls can sit unread for weeks.
Never publish a claim you cannot stand behind
The whole model depends on the engine trusting the page. One invented statistic costs more than it earns, which is why our drafting checks every number and quotation against what the customer actually supplied.
Should a small business pay for a press release in 2026?
Yes, if you are buying the right thing. No, if you are buying the old thing.
Do not pay for media coverage
In a month of tracking, not one visit to a release came from a news outlet, and I cannot point to a journalist who called because of one. If a service promises distribution to ten thousand outlets, what you are buying is your text on syndication pages that neither people nor AI engines treat as a source. The wires will charge you between $350 and $1,200 for that.
Pay for being the source an AI engine quotes
That is a real, measurable outcome and it arrives on a schedule. The last nine releases we published were all crawled within hours, and eight were cited within a week, most of them in the first position of the answer. At €49.95 a release, or €9.95 for the first one, that is the cheapest visibility signal a small company can buy, and the incumbents are not competing for it yet.
What to expect, honestly
- Where it works today is mostly Perplexity. As of October 2026, 64 of our last 71 citations came from Perplexity, 5 from Gemini and 2 from Claude. ChatGPT fetches releases when a user asks a relevant question, and we see those fetches, but it does not yet cite them routinely. Google AI Overviews cites our pricing guides heavily and our customer releases rarely. If your customers only ever use Google, the effect is weaker.
- Most citations answer brand questions, not buying questions. Thirty-eight of those 71 were for "X latest news" style queries, and six were for genuine questions. A release makes sure the engine says the right thing when someone asks about you. Getting recommended for a buying question takes a strong fact, an owned page that agrees with it, and some luck. It happened for PlantGift, but I would not promise it.
- Human traffic is small. Around four hundred human views across all releases in a month, against three and a half thousand AI fetches. The reader is a machine. Judge the release by what the machine says afterwards, not by the view counter.
Who should buy one
A new or small company with a concrete fact to announce, a price, a launch, a change, a number, in a niche where customers ask AI assistants what to buy. Shopify stores fit this exactly.
Who should not
A well-known brand whose own site already answers every question about it. A company with nothing specific to say. Anyone who needs phone calls from reporters or a traffic spike next week. A release will not do those things, and I would rather tell you now than take the €9.95.
My verdict
The press release is not dead. It has changed audience. Write it for the machine that answers the question, measure whether the machine quotes it, and it is the best money a small business can spend on visibility. Write it for a journalist who is not coming, and it is the worst.
If you want to see the schedule in more detail, we published the scan timings in How Long Until AI Cites Your Press Release? and the current prices on the pricing page.
If you have a fact worth announcing, write your first release for €9.95 and watch the scans yourself.