The state of HARO and reactive PR in 2026: 639 editions analysed

HARO was written off when Cision shut it down. Two years later it is back, free, and arriving three times every weekday. We received 639 editions between 2 January and 8 November 2026, about 13,400 journalist requests by our estimate. But the rules for getting quoted changed this year, and the changes hit health and wellness experts hardest.

Key findings:

• Half of HARO queries from June onward (50%) were labelled as refusing AI-written pitches, up from 38% in January to May.
• In health and pharma, 69% of queries from June onward asked for credentials or proof of expertise, up from 46%.
• Health and pharma made up about 1 in 5 HARO queries, peaking near 30% in April. Supplements and nutrition were the biggest health topic.
• HARO’s own data says journalists selected fully AI-written pitches about a third as often as human-written ones (vendor data).

What happened to HARO

HARO (Help a Reporter Out) started in 2008 as a free email of journalists’ requests for sources. Cision acquired it, rebranded it as Connectively in 2024, and then closed Connectively on 9 December 2024. In April 2025 Cision sold HARO to Featured.com, which brought it back as a free daily email funded by sponsor messages.

Since then, the platform has changed quickly:

• September 2025: HARO and Featured added Pangram AI detection, so journalists see a likelihood-of-AI score on every response.
• May–June 2026: Featured relaunched the Connectively name as its request platform, and launched a new Featured positioned as an AI assistant for PR. Connectively reported passing 100,000 users in June.
• September 2026: Connectively added automated “verification signals” to expert profiles, such as a professional-database match, a matching email domain and a LinkedIn profile. It says a typical journalist question draws 42 expert responses.

HARO, in other words, is no longer a free-for-all inbox. It is a filtered marketplace, and the filters are aimed at AI-written answers and unverifiable experts.

How we analysed 639 HARO editions

We counted every HARO edition delivered to one subscriber inbox from 2 January to 8 November 2026, month by month. We then took a sample of 34 digests, two to four per month across all three daily editions, and classified each of the 712 queries in them by HARO’s own category label, by topic, and by whether the query refused AI pitches or asked for credentials.

Download the monthly data: haro-2026-signals.csv (editions received, sampled queries, health share, no-AI labels and credential requests by month).

Each sampled digest held between 15 and 26 queries, an average of 20.9. Multiplying that by 639 editions gives our estimate of about 13,400 queries so far this year. The sample covers about 5% of editions and topics were tagged by keyword, so treat topic shares as directional. Comparisons between January–May (305 queries) and June–November (407 queries) are more reliable than month-to-month changes.

What journalists asked for: the category mix

Business and finance, mostly personal finance, insurance and housing, was the biggest category. Health and pharma was steady at about one query in five, rising from 9% of the sample in January to about 30% in April and 25% in October. Lifestyle grew in the second half, helped by beauty and gift-guide requests heading into the holidays.

The health topics reporters chased

Within the 137 health and pharma queries in our sample:

• Supplements, nutrition and diet came up in 41 (about 30%), the largest single topic.
• Mental health, stress and burnout came up in 26 (19%).
• Healthcare access and cost, such as Medicare, insurance, telehealth and prescriptions, came up in 18 (13%).
• Parenting (13), sleep (11), AI in health (11) and beauty (10) followed.
• GLP-1 and weight loss came up in 9. Across all categories, GLP-1 queries doubled from 1.0% to 2.2% of queries between the two halves of the year, from a low base.

That GLP-1 figure measures something different from our GLP-1 study, which found GLP-1 in 22% of editions from August to October. A single edition holds about 21 queries, so a topic can appear in one edition in five while accounting for only a couple of percent of queries. Both measures point the same way: rising, but from a small share.

AI as a story topic, such as AI in business or AI and jobs, held steady at about one query in nine all year. Tariffs were rarely named outright (6 queries), though cost-of-living, housing and insurance-cost questions together made up about one query in seven.

“No AI pitches”: the trust crackdown in numbers

The clearest change in 2026 was journalists shutting out AI-written answers.

By month, the no-AI label went from 35–42% in January to May to 55% in September and 57% in October. It is now the majority.

The platform itself changed tactics too. From January to mid-June, most queries in our sample (50–79% each month) carried a hidden, encoded instruction designed to make an AI-drafted reply give itself away. It is absent from every digest we sampled from 29 June onward. HARO has not said why, but the change appears to coincide with Pangram’s detection scores becoming the main defence.

HARO’s own data, published with Pangram in September 2026, puts numbers on the cost of using AI. Across 1.39 million pitches to 42,392 requests from June 2025 to July 2026:

• About 21% of pitches were fully AI-generated.
• Journalists selected about 1 in 10 fully AI pitches, against about 1 in 3 human-written ones.
• 35% of journalists switched on blocking of fully AI pitches, although it is off by default.
• Banned accounts were about 3.6 times as likely to have sent a fully AI pitch.

These are vendor figures, and “selected” means marked as useful on the platform, not published, so treat 1 in 3 as an upper bound. Qwoted has gone further: its terms prohibit AI-written pitches, and Press Gazette reported in July that it was removing more than 15 rule-breaking accounts a day.

Credentials are the new price of entry

In health, being quoted now depends on being checkable. More than two-thirds of the health and pharma queries we sampled from June onward asked for a licence, a board certification, a relevant degree, or a profile that proves the person exists and works in the field.

The reason is not abstract. In February 2026, Press Gazette reviewed 250 articles quoting experts in five UK tabloids and found 24 experts who did not exist or could not be verified, most of them supplied by PRs chasing links. Publishers responded with trusted-agency lists and stricter sourcing rules, and Connectively’s verification signals followed in September.

For health brands, this rewards the slow work: a named clinician or qualified expert, a consistent bio across the web, and answers that stay inside that person’s real expertise.

Fewer reporters, more reliance on PR

US media cut about 17,000 jobs in the first 11 months of 2025, and the cuts continued in 2026. Health and lifestyle titles were hit directly: Condé Nast closed Self as a standalone brand in April, and Good Housekeeping US had layoffs in August.

The journalists who remain lean on PR. In Muck Rack’s 2026 survey of 897 journalists, 86% said at least some of their stories come from pitches, and 88% said they ignore pitches outside their beat straight away. Cision’s 2026 survey of about 1,900 journalists found 66% rely on PR content for story ideas, while 53% oppose AI-generated pitches. Our reading: smaller newsrooms create more openings for well-sourced experts, and less patience for everything else.

Why a press mention matters more in the age of AI search

A press mention sends fewer clicks than it used to. Google referrals to publishers fell about a third worldwide in 2025, according to Chartbeat data in the Reuters Institute’s 2026 trends report, and Ahrefs measured a 58% drop in click-through for the top result when an AI Overview appears.

But a mention now does a different job. Muck Rack’s May 2026 analysis found about 84% of the links cited by ChatGPT, Claude and Gemini come from earned media, with journalism around a quarter of them. When people ask an AI assistant which supplement brand or clinic to trust, the answer is built from coverage. HARO’s sponsor messages noticed: from September 2026, they started selling earned media as the route to AI visibility.

What this means for health and wellness brands

• Answer as a real, named expert. Lead with the credential the journalist asked for, and link to a profile that proves it.
• Write it yourself. Half of queries refuse AI pitches, and the platforms can see AI-likelihood scores.
• Stay in your lane. Answer the queries that match the expert’s real field. Off-beat answers get ignored, and on Qwoted they can get an account removed.
• Be fast and specific. With dozens of responses per question, a short answer with one concrete fact or figure beats a long one.
• Count the coverage, not the clicks. The value of a placement is increasingly in being cited, by journalists and by AI assistants.

This is how our reactive PR for health experts works: we watch the requests, match them to a verified expert, write the answer with them, and charge USD 300 only when a placement goes live.

Method and limits

• Edition counts: every HARO edition received from [email protected], 2 January to 8 November 2026, counted by month.
• Query sample: 34 digests (12 morning, 16 afternoon, 6 evening), two to four per month, 712 queries. Categories use HARO’s own labels. Topics, no-AI labels and credential requests were tagged by keyword, and a query can carry more than one topic.
• Limits: one inbox, a roughly 5% sample, and keyword tagging. Monthly samples are small (41 to 92 queries). We report aggregate counts only and do not quote or identify any journalist’s request.
• Third-party data: figures from HARO/Featured, Pangram, Connectively, Qwoted, Muck Rack, Cision and Ahrefs come from companies that sell products in this market. We have labelled them as vendor data in the sources.