25 AI Prompts for PR and Media Outreach

TL;DR: These 25 prompts cover the four things PR and comms people write on repeat: pitches, boilerplate, expert commentary, and HARO-style responses. Paste them into ChatGPT, Claude, or Prompt Architects, fill in your specifics, and always fact-check anything the model produces before it reaches a journalist's inbox.

Most PR work is not creative writing. It's fast, repeatable drafting under a deadline: a pitch that has to go out in the next twenty minutes, a boilerplate paragraph that needs updating for the fifth time this quarter, an expert quote for a reporter who needs it by end of day. That's exactly the kind of writing AI is good at, provided you feed it real specifics and don't let it invent facts, names, or statistics on your behalf.

This matters more in PR than in most other kinds of business writing, because the output here doesn't just sit in your own inbox or CMS. A pitch, a quote, or a boilerplate paragraph gets read, judged, and sometimes reprinted verbatim by someone outside your organization who has no way of knowing which parts were drafted by a model and which parts you verified yourself. That external audience is the reason the discipline in this post leans harder on verification than a typical prompt-list post would.

The 25 prompts below are organized into four sections that map to how PR work actually gets done: pitching journalists, writing and updating boilerplate, producing expert commentary, and answering HARO-style reporter queries. Copy any of them, fill in the bracketed variables, and treat the output as a first draft, not a final one.

What makes a PR prompt actually work?

A PR prompt works when it forces you to supply the one thing that makes a pitch or quote non-generic: a real, specific detail about the journalist, the news moment, or your own data. A prompt that says "write me a pitch about my company" produces exactly the kind of interchangeable pitch that gets deleted unread. A prompt that says "write me a pitch referencing [journalist]'s piece on [specific recent story] and connecting it to [specific data point]" produces something a person could plausibly have written by hand.

Every prompt below has at least one required input field for this reason. If you skip filling it in and let the model guess, you'll get filler.

There's also a structural reason so much PR outreach reads the same: most people write a pitch, a boilerplate, and a HARO response as three unrelated tasks, starting from a blank page each time. Treating them as three variations on the same underlying discipline, feed the model real facts, ask for a specific structure, and cut anything that could apply to any company, produces noticeably tighter drafts across all three, because the same failure mode (genericness from missing specifics) shows up in each one.

How PR outreach formats differ

Before jumping into the prompts, it helps to see how the four formats in this post actually differ in what they're optimized for. A pitch is judged almost entirely on whether the first two sentences earn a reply; boilerplate is judged on accuracy and consistency across dozens of reuses; expert commentary is judged on whether it sounds like a real, specific opinion rather than a safe generality; and a HARO-style response is judged on speed and quotability, since reporters are often working against a same-day deadline and skimming dozens of responses.

Knowing which of these four you're writing before you open a prompt changes what "good" looks like. A pitch that reads like boilerplate gets skipped; a boilerplate that reads like a pitch sounds unprofessional when it's quoted verbatim in a news story; and expert commentary that hedges every sentence with "it depends" or "generally speaking" reads as unquotable no matter how accurate it is. Each format has a different job, and using the wrong register for the job is a more common mistake than poor grammar or weak phrasing.

Pitch prompts (1-9)

These are for cold and warm pitches to journalists, freelancers, and newsletter writers.

1. The news-hook pitch

Write a PR pitch email to a journalist named {{journalist_name}} who covers
{{beat}} at {{outlet}}. Reference their recent article "{{recent_article_title}}"
in the opening line, then connect it to this story: {{your_story_summary}}.
Keep it under 130 words, three short paragraphs, no subject line yet.
Do not invent any statistics or quotes I have not given you.

2. The subject-line generator

Write 8 subject lines for a pitch about {{story_summary}} going to a
journalist who covers {{beat}}. Keep each under 60 characters. Avoid
clickbait phrasing, exclamation points, and words like "revolutionary"
or "game-changing". Make at least 3 lead with a number or a proper noun.

3. The exclusive-offer pitch

Write a pitch offering {{journalist_name}} a 48-hour exclusive on this story:
{{story_summary}}. State the exclusivity window and the embargo date clearly.
Explain in one sentence why this specific journalist is a good fit based on
their coverage of {{their_beat_or_prior_story}}.

4. The follow-up pitch (no reply)

Write a one-paragraph follow-up to a pitch I sent {{days_since}} days ago
about {{story_summary}}. Do not repeat the full original pitch. Add one new
piece of information: {{new_detail}}. End with a low-pressure close that
makes it easy to say no.

5. The data-story pitch

I have this data: {{data_points}}. Write a pitch to a journalist covering
{{beat}} framing this data as a story, not a stat dump. Lead with the most
surprising number. Suggest one angle the journalist could take that isn't
just "here's our data".

6. The newsjacking pitch

A current news story is: {{current_news_summary}}. My company/expert has
this relevant angle: {{your_angle}}. Write a same-day pitch connecting the
two for a journalist covering {{beat}}. Flag if the connection feels forced
rather than genuine, and suggest a stronger angle if so.

7. The multi-outlet pitch variation

Take this pitch: {{base_pitch}}. Rewrite it in 3 versions: one for a trade
publication covering {{trade_beat}}, one for a general business outlet, and
one for a newsletter writer with a personal, informal voice. Keep the core
facts identical across all three.

8. The founder-story pitch

Write a pitch positioning {{founder_name}}, {{title}} at {{company}}, as a
source for stories about {{topic}}. Include 2 sentences of relevant
background and one specific, non-obvious opinion they hold on the topic.
Do not use the words "passionate" or "thought leader".

9. The podcast-guest pitch

Write a pitch offering {{guest_name}} as a podcast guest for shows covering
{{topic}}. Include 3 specific talking points they could discuss, one of
which should be contrarian or surprising rather than a standard talking point.

Boilerplate and company description prompts (10-14)

10. The standard boilerplate

Write a 50-word company boilerplate for {{company_name}}, a {{one_line_
description}}. Include what the company does, who it's for, and one
concrete differentiator. Do not use "innovative", "cutting-edge", or
"leading provider".

11. The boilerplate refresh

Here is our current boilerplate: {{current_boilerplate}}. Update it to
reflect this new fact: {{new_fact}}. Keep the length within 10 words of
the original and preserve any phrasing that's become recognizable.

12. The audience-specific boilerplate

Take this boilerplate: {{boilerplate}}. Write two variants: one for a
technical trade audience that can include jargon like {{technical_terms}},
and one for a general consumer audience that avoids all of it.

13. The executive bio

Write a 75-word bio for {{name}}, {{title}} at {{company}}. Include:
{{background_facts}}. Lead with their current role, not their history.
End with one specific credential or achievement, not a generic claim
about expertise.

14. The one-liner for a media kit

Compress this into a single 15-word sentence for a media kit header:
{{longer_description}}. It should work standalone with no other context.

Expert commentary and quote prompts (15-20)

15. The reactive quote

A reporter is writing about {{news_event}} and wants a quote from
{{expert_name}}, who knows about {{relevant_expertise}}. Draft a 2-sentence
quote reacting to this news, in the expert's likely voice. Flag any part
of the quote that would need to be fact-checked against real data before
sending.

16. The trend-commentary quote

Write 3 alternative quotes from {{expert_name}} commenting on the trend of
{{trend_description}}. Each should take a genuinely different angle
(supportive, skeptical, and nuanced), not just reword the same opinion.

17. The op-ed opening

Write a 100-word opening for an op-ed by {{author_name}} arguing that
{{thesis}}. Open with a specific scene, statistic, or anecdote rather than
a general statement. I will supply the supporting statistic separately —
do not invent one.

18. The panel talking points

{{expert_name}} is speaking on a panel about {{topic}}. Draft 5 talking
points they could use, each phrased as a complete, quotable sentence
rather than a bullet fragment. Include at least one point that pushes
back on the panel's likely consensus view.

19. The "3 predictions" commentary piece

Write 3 predictions from {{expert_name}} about {{industry/topic}} for
{{year}}, each with a one-sentence justification. Make each prediction
specific and falsifiable, not vague statements like "AI will keep growing".

20. The rebuttal quote

A competitor or critic said: "{{quote_to_respond_to}}". Draft a measured,
non-defensive response quote from {{spokesperson_name}} that addresses the
substance without sounding combative or dismissive.

HARO-style response prompts (21-25)

21. The tight-answer response

A reporter asked this query: "{{query_text}}". Write a 60-word response
from {{expert_name}}, {{title}} at {{company}}, that directly answers the
question in the first sentence. End with a one-line credential.

22. The multi-angle response

A reporter asked: "{{query_text}}". I have relevant experience in
{{background}}. Draft 2 different response angles: one focused on data
and trends, one focused on a specific personal or company anecdote.

23. The credential-first response

Write a HARO-style response opening with {{expert_name}}'s single most
relevant credential for this query: "{{query_text}}", then answering in
2 sentences. The credential should be the actual hook, not an afterthought.

24. The declining-gracefully response

A reporter's query "{{query_text}}" is outside our real expertise. Draft a
short, polite response declining to comment but suggesting one specific
type of source they should look for instead, without naming a competitor
by name.

25. The batch-query triage prompt

Here are 5 reporter queries: {{list_of_queries}}. For each, tell me in one
line whether {{expert_name}} has genuine relevant expertise to respond, and
rank them by how strong a fit they are. Do not recommend responding to a
query just because it's easy to answer.

Adapting one pitch without sounding copy-pasted

A common shortcut is writing one pitch and sending a near-identical version to twenty journalists on the same beat. Reporters notice this, and a pitch that's obviously been blasted to a whole beat list tends to get a much colder reception than one that reads as though it was written for the person receiving it. The fix isn't writing twenty pitches from scratch, it's using AI to genuinely vary the framing, not just swap the name field.

Concretely, that means giving the model each journalist's actual recent coverage and asking it to change the angle, not just the greeting. A journalist who recently covered the funding side of your industry gets a version emphasizing your company's traction; one who covers product trends gets a version emphasizing what's actually new about the product. Prompt 7 above (the multi-outlet pitch variation) is built for exactly this, but it only works if you supply real, distinct context for each version. Feed it the same generic description three times and you'll get three generic pitches with different headers.

This is also where a saved prompt library earns its keep. If you're pitching the same story to a dozen outlets over a week, keeping the base facts, the story summary, and the data points in one place, and reusing them as variables rather than retyping them for each journalist, cuts both the drafting time and the risk of a fact drifting between versions.

Where AI genuinely helps, and where it doesn't

AI is fast at the parts of PR that are structural: getting a pitch into three clean paragraphs, compressing a bio to 75 words, generating subject-line variations, and reformatting the same core message for different outlets. It is not good at the part that actually gets a pitch opened, which is genuine, specific knowledge of what a particular journalist covers and cares about right now. No prompt substitutes for reading the journalist's last five articles before you pitch them, and no amount of prompt engineering turns a vague, one-line brief into a pitch that reads as personally researched.

The same caution applies to quotes and commentary. A model will happily produce a confident-sounding statistic or a specific-sounding fact if you don't supply one, phrased with exactly the same fluency as a fact you did supply. That's the trap: there's no visible difference in tone between a verified number and an invented one in the model's output, so the discipline has to come from you, checking every figure before it leaves your draft. In PR specifically, an invented number in a quote that gets published is far more damaging than a slow response, because it's now attributed to a real, named person in print, and retracting a misquote after publication is a much harder conversation than asking for one more day to get the pitch right.

Treat every AI-drafted quote or statistic as a draft that needs the expert's actual sign-off, not a finished deliverable. The same applies to boilerplate: a compressed 50-word paragraph that gets copy-pasted into dozens of stories over a year deserves a careful read before it first goes out, since a small factual error at that stage propagates every time it's reused. None of this means avoiding AI for PR work; it means treating it the way a good junior staffer's first draft gets treated, useful, fast, and always reviewed before it goes external.

Tracking what actually works

One habit worth building alongside these prompts: keep a simple log of which pitches got a reply and which didn't, along with what angle each one used. After a few dozen pitches, patterns show up that no prompt can predict in advance, some journalists respond better to data-led pitches, others to founder-story angles, and the same outlet can behave differently depending on which section is covering your beat that quarter. Feed that log back into your prompts over time (prompt 2's subject-line generator and prompt 7's multi-outlet variation both improve noticeably once you tell the model which angles have actually landed before) and the pitches get sharper without you having to reinvent the format each cycle.

This is also a reasonable way to decide when a beat or outlet isn't worth continued effort. If ten pitches to the same publication with genuinely varied angles all go unanswered, that's more useful signal than any single response would be, and it's worth reallocating the time toward outlets where your angle is actually landing rather than continuing to refine pitches nobody there is reading.

If you're running these prompts across a full comms calendar — pitches this week, boilerplate updates next month, HARO responses daily — a prompt manager that saves the ones that actually land and lets you swap in new variables per journalist saves more time than rewriting a fresh pitch prompt every time. For a related workflow built on the same specificity-first approach, see 30 AI Prompts for Cold Outreach and Outbound Email.