Discussion starter: a digital PR question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Digital PR
HARO-style expert commentary platforms are flooded with AI pitches. Is reactive expert commentary still worth the time for links?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Reactive expert commentary can still be worth the time, but as a channel for credible mentions with links as a by-product, not as a link-volume channel. The flood of AI-written pitches has not changed what journalists need. It has changed how many responses they wade through to find it, and that pushes the value towards experts with checkable credentials, specific experience and something a chatbot cannot produce. If your process is a junior (or a model) answering dozens of queries a week in a generic voice, the economics have probably already broken. If a real expert answers a few carefully chosen requests with data or first-hand detail, it can still earn links and mentions that are hard to get any other way.
The number that decides it is the expert’s time per placement you would actually want, not the number of pitches you can send.
What the public data says about request quality
The original HARO was rebranded as Connectively, shut down, and then relaunched with a new CEO, Brett Farmiloe. In a quote carried in BuzzStream’s study of journalist request platforms, he said “Quality is our top priority at HARO, which is why every request is manually reviewed before approval”. Request volume has spread across Qwoted, Featured, Source of Sources and social hashtags such as #journorequest.
That BuzzStream study (published July 2025, using data compiled by PressPulse.ai over May and June 2025, with authority measured by Ahrefs Domain Rating) is the most useful public snapshot of what these platforms actually offer:
Finding What it implies for link planning Only 21.9% of requests explicitly offered follow links Most placements come with no stated link terms. Plan on mentions, and treat followed links as upside. Just over 38% of HARO’s requests came from sites in the DR 0 to 9 bucket A large share of requests on some platforms come from small or new sites, not established publications. 70.3% of Qwoted’s opportunities came from DR 70 to 100 sites Platforms differ a lot. Where you look matters more than how many requests you answer. 36.5% of Featured’s requests were in marketing and PR Some platforms lean heavily towards industry content sites, where the reader is not your customer. Keep the caveats in view. Domain Rating is a third-party metric, not a Google signal. “Explicitly offered” means many requests simply did not state link terms either way. The data is a two-month snapshot from mid-2025. It says nothing about how many pitches each request receives or how many of them are AI-written, so it describes the supply of opportunities, not your odds of winning one.
Why the AI flood changes the maths
Journalist attention is fixed. When the number of responses per request climbs, the acceptance rate per pitch falls, and the journalist filters harder on signals that are expensive to fake. Think of a job ad that suddenly draws 500 applications instead of 50. The hiring manager does not read each one more carefully. They skim for the few things that separate a real candidate from a template: a specific achievement, a credential they can check, evidence the person read the ad.
AI-written responses are fluent but converge on the same points, because they are drawing on the same general knowledge. A response that sounds like the average answer becomes invisible, even when a human wrote it. What still stands out is material a model cannot generate for you:
- a number from your own work (jobs completed, average project cost, return rates, claim volumes)
- a concrete, anonymised case with a detail only a practitioner would know
- a clear position with a reason, including disagreeing with the premise of the request
- a named person with a credential the journalist can verify in under a minute
- availability for a quick call or follow-up question
This is reasoning from how editorial sourcing works rather than a measured result, but it explains why the same channel can be dead for one business and productive for another.
Three outcomes that get bundled together as “links”
Followed link in a relevant publication
The outcome most programmes are built around, and the rarest. It only counts if the article is indexed and the publication is one your audience or industry recognises.
Named mention, nofollow or no link
Still useful for credibility, sales material, “as featured in” trust, and for journalists coming back to you. Whether mentions influence AI answers is plausible but not well measured. Google’s guide to generative AI features says seeking inauthentic mentions is not helpful, which rules out volume-chasing as a strategy.
Link from an expert roundup or content site
Cheap and plentiful. A page that publishes 40 expert answers, each with a link, is built to attract contributors rather than readers. The likelier outcome is that such links do little, not that they cause harm, but they are not worth an expert’s hour.
When it is still worth the time
Condition Lean towards doing it Lean towards stopping or shifting Expert availability A named expert can give a few hours a week and take calls Responses are ghostwritten by someone who cannot answer follow-ups Type of expertise Regulated or specialised fields (finance, law, health, trades, engineering) where journalists must verify sources Generalist tips that anyone, or any model, could offer Proof Your own data, real cases, credentials shown on an expert bio page Opinion without evidence Where requests come from Publications your customers read Mostly roundups and marketing content sites Goal Authority, mentions and journalist relationships, with some links A fixed monthly quota of followed links Competition Niche or regional topics with few credible respondents Crowded national consumer topics How I would run it now
- Filter before writing anything. Only answer requests from outlets you would be glad to list on your site, on topics where the expert has direct experience. Skip anonymous requests and obvious roundups.
- Put the element a model cannot produce in the first two sentences. The number, the case detail or the clear position goes first, because journalists skim.
- Keep it short and quotable, with name, title and credential, and a link to an expert bio page that proves them (qualifications, prior coverage, professional profiles). That page also gives journalists something to link to.
- Use AI for triage if it helps, not for substance. Sorting requests and flagging relevant ones is a reasonable use. If a journalist could get your answer by asking a chatbot, the pitch has no reason to exist.
- Offer a call and answer follow-ups quickly. Being reachable is a filter that template responders fail.
- Turn placements into relationships. Thank the journalist, share the piece, offer to be a source on their beat, and pitch proactively when news breaks in your field. Proactive commentary on breaking news often beats reactive requests because you are not competing inside a queue.
Measure it like a channel, with a stop rule
Track each request at placement level: request source, outlet, pitch sent, used or not, link status (followed, nofollow, none), whether the article is indexed, referral sessions, and any repeat contact from the journalist. The metric that matters is expert hours per placement you would want, compared with what those hours could produce elsewhere.
Hypothetical example:
An expert spends 3 hours a week for 12 weeks (36 hours) and answers 40 filtered requests. Six responses are used, three include links, and two of those are followed links on publications relevant to the business. That is 18 expert hours per followed relevant link, plus four named mentions and, say, two journalists who ask for comment again. Whether 18 hours is good depends entirely on what else that expert’s time could produce, such as a data study, a proactive newsjacking pitch, or client work. These figures are illustrative, not benchmarks.
Set the decision rule before you start. After 12 weeks, if hours per relevant placement are clearly worse than your other link sources and no journalist has come back to you, stop reactive responses and move the time into proactive commentary or data-led PR. If journalists are returning with direct requests, the per-link cost overstates the true cost, because repeat sourcing is where this channel compounds. Also check which platforms produced the placements. On the BuzzStream data, the spread between platforms is wide enough that dropping one and doubling down on another can change the result more than writing better pitches.
Watch for this:
Paying for “guaranteed” expert placements, or responding at scale with keyword-rich anchor text requests, turns editorial sourcing into something closer to a link scheme. The value of a quote comes from the journalist choosing it. Anything that removes that choice removes the value.
My position: keep reactive commentary if you have a genuine expert, a small time budget and the discipline to answer only requests that fit. Judge it on relevant placements and relationships, not raw link counts. If the only way to hit your numbers is volume, the AI flood has already made that version of the channel unprofitable, and the time is better spent on proactive stories built on your own data.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.