Pitching Journalists With AI: Press Releases, Story Angles, and Follow-Up

What Earned Media Actually Delivers

Be honest about the mechanism before investing in it. A news story is unlikely to drive a wave of donations, and organizations that pursue coverage expecting a direct fundraising return usually abandon media work within a year. The value is real but indirect. The first thing coverage buys is third party credibility, and it matters most in rooms where you are not present: a program officer doing due diligence, a county commissioner weighing a contract, a prospective board member checking whether you are real.

The second thing it buys is reach into audiences you have no other route to. Your email list is people who already know you. Your social following largely agrees with you already. A local story reaches people who were not looking, including the ones who need your services and do not know you exist. For direct service organizations, coverage is frequently the most effective client outreach available, and it is underrated precisely because communications planning treats media as a fundraising activity.

The third is agenda setting. If you work on housing, evictions, food insecurity, or any issue where decisions get made by people reading local news, coverage is how your framing enters the conversation. You are supplying the vocabulary and the numbers a council member repeats three weeks later. Set against the alternatives the tradeoff is clean: paid media buys control and timing with money you would rather spend on programs, owned channels give you full control over an audience you already have, and earned media delivers credibility and reach that neither can manufacture. Know which one you are doing and stop expecting one to do another's job.

Do You Actually Have a Story? The Step Everyone Skips

This is the decisive step, and it is the one nonprofits skip because it requires saying no to something the board is excited about. A news story needs a hook, which is the reason this information is worth publishing now, to this outlet's readers, rather than at any other time or not at all. Without a hook you do not have a pitch, you have an announcement, and announcements are what fill the deleted folder.

There are a limited number of genuine hooks. Data nobody else holds is the strongest available to nonprofits and the most consistently wasted: how many families you turned away last month, what your waitlist looks like against a year ago, which zip codes your intake calls come from. That is original reporting a journalist cannot get elsewhere, sitting unexamined in your case management system. A genuine first counts, if it is first in the field rather than first for you. A local angle on a national story is the most reliable hook in local media, because when a federal program changes the reporter needs someone locally who can say what it means on the ground, and being the organization that answers the phone that day beats a year of press releases.

Rounding out the list: a deadline or a date creates urgency, whether a funding cliff, a legislative vote, or an enrollment period closing. An unexpected finding works because it violates an assumption, so the counterintuitive result in your evaluation is more publishable than a confirmation of what everyone believed. A person whose story carries it makes the others human, provided you have genuine informed consent and the person understands what publication means. That last condition is not optional, and it is the one most often handled badly under deadline pressure.

Now the harder half. Your gala is not a story. Your new development director is not a story, though the trade press may run a personnel note. Your rebrand is not a story. Your thirtieth anniversary is not a story, although the thirty years of data behind it might be. A grant is not a story unless it is unusual in size or source. None of this means these things do not matter. It means they matter to you. A useful test is to write the headline you imagine the outlet running, in their style, and ask whether a stranger would click it. AI helps here: ask a model to argue the other side and explain why a reporter would pass, and that critique is often worth more than the draft.

Writing a Pitch a Reporter Will Actually Open

Email is overwhelmingly the channel journalists want pitches in, and the reporter you are writing to is receiving a great many. Propel's Media Barometer, which aggregates behavioral data from hundreds of thousands of real pitches, consistently finds response rates in the low single digits, with shorter pitches outperforming longer ones by a wide margin. Cision's survey work points at the reason: the overwhelming majority of journalists say they reject a pitch immediately if it is not aligned with their beat or audience. Length and relevance are doing nearly all the work.

Treat the subject line as a headline, not a label. "Press Release: Local Nonprofit Announces New Program" tells a reporter nothing except that this is not urgent. "Eviction filings in Marion County up 31% since January, we have the intake data" tells them what the story is, that there is a number in it, and that you have something they cannot get elsewhere. Keep it short enough to survive a phone screen and make it specific. The same discipline you would apply to donor email applies here, and our guide to writing subject lines with AI transfers directly, with the caveat that a journalist is even less patient than a donor.

The first two sentences determine everything. Sentence one states the story, not your organization. Sentence two says why it belongs to this reporter and this outlet right now, ideally by referencing something specific they wrote. Do not open with "I hope this finds you well," your mission statement, or your founding year. A reporter reading on a phone between assignments decides inside those two sentences, and anything placed before the story pushes the story below the point where they stop.

Then keep it short. Muck Rack's annual State of Journalism work has repeatedly found that roughly two thirds of journalists prefer pitches under 200 words, and one journalist quoted there put it bluntly: reporters seldom read past 200 words at all. A good pitch is the story in one sentence, the evidence in two or three, what you can supply in one, and an offer to talk. Everything else goes in a fact sheet you mention and send on request. Do not attach anything, because attachments trip spam filters and do not open on phones. Mention that a press release exists rather than leading with it: a release is a formal document written for the record, while a pitch is a personal note explaining why a story is one person's. Sending the release instead of the pitch is the most common error in nonprofit media outreach, and it is why so much of it reads as addressed to nobody.

Using AI Well: The Tasks Where It Genuinely Helps

AI helps in media relations when it is pointed at structure, preparation, and volume of options rather than at the final artifact sent to a person. The distinction that matters is between work done on your own facts and work that invents facts, and between work that prepares a human and work that replaces one.

Generating multiple angles from one set of facts. This is the highest-value use and the one most organizations never try. Give a model your program numbers, what changed, and who is affected, then ask for six story angles aimed at six outlet types: the data angle for the regional daily, the human angle for local television, the policy angle for the trade publication, the community angle for the neighborhood newsletter, the sector angle for a philanthropy outlet, the conversational angle for a podcast. Most nonprofits pitch one angle everywhere and conclude the story was not interesting. Usually the facts were fine and the framing was wrong for five of the six.

Tailoring a pitch to a specific reporter. Paste in three of a reporter's recent articles and ask what they consistently care about, what evidence they use, how they frame the issue, and what question your story would answer for them. Then draft against that. This is what converts a generic pitch into one that demonstrates you read their work, which is the biggest determinant of a reply. Use real articles you actually retrieved, never the model's recollection of what a reporter covers.

Writing quote options for a person to react to. Executive directors freeze when asked for a quote in the abstract. Giving them three drafted options and asking which is closest to true produces a better and faster result than a blank request, provided the rule is absolute: nothing is attributed to a person until that person has read it and approved it in writing. A documented approach to brand voice consistency pays off here, because the drafts come back sounding like your organization rather than a press release generator.

Building the materials you never get around to. A one-page fact sheet with your key numbers, service area, history, and leadership. A boilerplate paragraph you actually like. A glossary of terms specific to your field. A backgrounder on the policy context. These make a reporter's job easier, and they are why some organizations get called back and others do not. All are straightforward drafting tasks from material you already have.

Interview preparation and research summarizing. Ask for the twenty questions a skeptical reporter would put to you, including the uncomfortable ones about cost per participant, outcomes you cannot prove, and why the problem has not improved, then have your spokesperson answer them out loud. A model can role-play the interview itself, which is effective preparation for someone who has never done one. It can also find the publishable finding inside your evaluation report and express it in a sentence a reporter can print. That is scaling storytelling with AI applied to a harder audience, with an identical constraint: the model organizes what is genuinely in your data, and the moment it supplies numbers of its own you have left the useful zone.

Follow-Up, Embargoes, and Exclusives

Follow-up is where anxious organizations destroy good work. The convention journalists describe is easy to follow: one follow-up, sent three to seven days after the original, and then nothing. About half of journalists surveyed by Muck Rack say a single follow-up is right, and a similar share put the timing in that window. Following up within hours reads as pressure. Following up three times reads as someone who will keep emailing regardless of the answer, which is the reputation that gets you filtered.

Make the follow-up add something. "Just checking in on the below" asks the reporter to do work. A note saying the number has moved, that a new development makes the story timelier, or that you can now offer a specific interview gives them a fresh reason to look. Keep it to three sentences. When there is no response to that, the answer is no, which is usually not a judgment about your organization but a reporter who had four other things that day. Log it and move on, because the same person may say yes in four months.

An embargo is an agreement that you will give a reporter information in advance on the condition that they do not publish before a stated time. It exists so a journalist can report properly rather than rushing, and so several outlets can publish at once. Three things make it work. It must be agreed in advance, not declared: writing "embargoed until Tuesday" atop an unsolicited email creates no obligation, because a reporter who did not agree is not bound. The time must be specific, including the time zone. And it is only worth using when advance time genuinely helps the reporting, which for most nonprofit news is rarely. Using one unnecessarily makes you look like you are performing importance.

An exclusive is the other side of that coin: you offer one outlet the story first, and they commit to covering it with more attention. It can be an excellent trade, because one well-reported feature beats five thin mentions and the piece that runs first shapes how everyone else covers it. The cost is that you are betting on one outlet, and if they pass after two weeks your timing may be gone. Be explicit about the window, and never offer the same exclusive twice, because outlets talk and that is unrecoverable. The same instinct governs the rest of the etiquette: do not ask to preview a story, do not seek quote approval beyond confirming a direct quotation's accuracy, and never contact an editor because a reporter did not reply.

Measuring Earned Media Without Fooling Yourself

Media measurement in the nonprofit sector is frequently dishonest, usually not deliberately. The standard board report shows a large impressions figure, an advertising value equivalency, and a count of mentions, and none describe anything that happened. Impressions estimate how many people could theoretically have seen an outlet, not how many read your paragraph. Advertising value equivalency prices coverage you did not buy and could not have bought, and the public relations standards bodies rejected it years ago. Reporting it trains your board to expect a number that means nothing.

Measure fewer things and make them real. Count placements and record quality alongside quantity: were you the subject or a passing mention, did your framing survive into the piece, does the outlet reach an audience you care about. Track message penetration by checking whether the points you prepared actually appeared, which tells you whether your preparation worked. Track referral traffic, which is small and honest. Track inbound contacts that mention the story, including the client who called because they saw it, which is often the outcome that mattered most. And track relationship depth as its own metric, because it leads everything else: how many journalists know who you are, how many have quoted you twice, how many contacted you rather than the reverse.

For finding coverage in the first place, free search alerts on your organization, your leaders, and your issue terms catch most of it. Monitoring tools make this easier and are worth understanding before buying anything, and our overview of AI-powered media monitoring covers where automation genuinely helps and where an alert does the same job for nothing. Set expectations with your board in advance, because that conversation prevents the quiet abandonment of media work after a slow year. A reasonable first-year goal from a standing start is a working media list, three or four relationships, and two or three placements. That is a genuine result, and far better than an impressive impressions figure that collapses when someone asks what it means.