How to Use Claude for PR: The 2026 Guide for PR Agencies and PR Professionals

Claude has become one of the most practical AI tools for PR, not because it writes your press release for you, but because it reads everything you never have time to read, drafts the first 70% of the work, and pokes holes in your messaging before a journalist does.

The timing matters. Between September 22 and October 7, 2026, Anthropic released a full new model family, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5, alongside Claude Fable 5.1, its most capable publicly available model. For PR agencies, the question is no longer whether to use Claude. It’s which Claude model to use for which PR job, and how to build it into the daily rhythm of pitching, client service, crisis response and reporting.

This guide answers exactly that, with a model-by-task matrix, a day-in-the-life workflow, an agency setup checklist and 10 copy-ready Claude prompts for PR.

Key takeaways

• Sonnet 5.5 is the everyday PR workhorse: press releases, pitches, bylines, briefing docs.
• Haiku 5.5 handles high-volume, easy-to-check work: coverage summaries, tagging mentions, cleaning media lists.
• Opus 5.5 is for strategy and anything high-stakes: crisis statements, campaign plans, client reports, new business.
• Fable 5.1 is reserved for the biggest, longest research and reporting projects.
• Set up one Claude Project per client with brand voice, key messages and boilerplate, and every draft starts on-brand.
• Claude drafts; PR professionals verify, decide and own the relationship. Never send raw AI output to a journalist.

In this guide

1. Why PR teams are using Claude in 2026
2. The Claude models explained for PR professionals
3. Which Claude model to use for each PR task
4. A day in a PR agency with Claude
5. How to set up Claude for a PR agency
6. 10 copy-ready Claude prompts for PR
7. Using Claude for crisis communications
8. Claude, AI search and the new earned-media reality
9. Guardrails: accuracy, confidentiality and ethics
10. FAQ

Why PR teams are using Claude in 2026

AI in public relations is no longer an experiment. Muck Rack’s State of AI in PR 2026 report, based on a survey of 564 PR professionals in December 2025, found that generative AI adoption has leveled off at 76%. Almost everyone who can use it already does.

Two other findings matter more for how you work:

• 98% of PR pros always or often edit AI-generated output. AI is a drafting partner, not a publisher.
• Only 12% use AI agents, and just 7% are comfortable letting agents take action on their behalf. Most teams are still using AI for one-off chats rather than real workflows.

That’s the opportunity. When adoption plateaus, the advantage shifts from using AI to using it well: choosing the right model, giving it the right context, and building it into repeatable agency processes. Claude is particularly suited to PR because so much of the job is reading long material (transcripts, reports, coverage, briefs) and producing polished writing in a specific brand voice.

The Claude models explained for PR professionals

Anthropic sells Claude in tiers. Bigger models think harder and handle longer, messier work; smaller ones are faster and cheaper. Here’s the current lineup translated into PR terms.

Model	Released	Think of it as	Best PR jobs	Speed / cost
Claude Haiku 5.5	Oct 7, 2026	The fast junior account executive	Coverage summaries, tagging and sorting mentions, media list cleanup, quick rewrites, social copy variants	Fastest · lowest cost
Claude Sonnet 5.5	Sep 28, 2026	The reliable senior account manager	Press releases, pitches, bylines, interview briefs, client emails, Q&A documents	Fast · mid cost
Claude Opus 5.5	Sep 22, 2026	The account director / strategist	Comms strategy, message houses, crisis plans, campaign proposals, monthly and quarterly reports, decks	Slower · premium
Claude Fable 5.1	Sep 1, 2026	The specialist you bring in for the hardest projects	Large reputation audits, multi-market launch plans, annual reporting across huge document sets	Slowest · highest cost

Claude Haiku 5.5: speed for high-volume PR tasks

Haiku 5.5 is the smallest and fastest member of the Claude 5.5 family, built for high-volume, cost-sensitive work like summaries and classification. In a PR agency, that means the repetitive tasks that eat hours: summarizing 60 overnight clips into a morning briefing, labeling mentions by sentiment and message pull-through, de-duplicating a media list, or producing ten variations of a social post. These are jobs where you can quickly check the output, so speed beats depth.

Claude Sonnet 5.5: the everyday PR workhorse

If you only pick one model, pick Sonnet 5.5. Anthropic positions it as the balanced default, and independent benchmarks put it close to Opus 5.5 on many tasks while running faster and cheaper. It’s the right choice for most daily writing: first drafts of press releases, journalist pitches, bylined articles, executive bios, spokesperson briefing notes and client update emails.

Claude Opus 5.5: strategy, crisis and client-ready deliverables

Opus 5.5 is Anthropic’s flagship for demanding reasoning and knowledge work. Anthropic says it performs at roughly the level of Fable 5.1 on most work while costing about 40% less to run than its predecessor, and that it writes more clearly and succinctly than earlier top models. It’s especially strong at turning messy information into finished business documents and presentations. In PR, use it whenever a weak answer would cost you: crisis holding statements, campaign strategy, message houses, quarterly coverage reports, award entries and new business proposals. Note that Opus 5.5 is not included on Claude’s Free plan.

Claude Fable 5.1: for the biggest projects only

Fable 5.1 belongs to Anthropic’s Mythos tier, which sits above Opus, and it’s the most capable Claude model available to the public. It shines on long, multi-step research and on document, spreadsheet and slide work that must stay coherent across a huge volume of material. It’s also slower and roughly twice the API price of Opus-class models. For PR, think of it for projects like: “Analyze two years of coverage across eight markets, compare it with three competitors and build the annual reputation report.” For everyday work, Opus 5.5 or Sonnet 5.5 will be faster and just as good.

What about Claude Mythos 5.1? It shares Fable 5.1’s underlying model but is restricted to a small group of vetted organizations under Anthropic’s Project Glasswing. PR teams won’t need it.

The simple routing rule

Start with Sonnet 5.5. Drop to Haiku 5.5 when the task is repetitive and easy to check. Move up to Opus 5.5 when a weak answer could cost you a client, a story or a reputation. Reserve Fable 5.1 for the projects even Opus struggles with.

Which Claude model to use for each PR task

Bookmark this matrix and share it with your team. It’s the fastest way to stop overpaying for simple work and underpowering important work.

PR task	Recommended model	Why
Morning coverage and news briefing	Haiku 5.5	High volume, easy to check, needed fast
Tagging mentions by sentiment, outlet tier, message pull-through	Haiku 5.5	Classification at scale
Cleaning and tiering a media list	Haiku 5.5 → Sonnet 5.5	Haiku sorts; Sonnet writes the angle notes per journalist
Press release first draft	Sonnet 5.5	Quality writing at speed
Personalized journalist pitches	Sonnet 5.5	Tone and tailoring for each reporter
Bylined articles and thought leadership	Sonnet 5.5 (Opus 5.5 for flagship pieces)	Voice consistency over long text
Spokesperson interview prep and mock Q&A	Sonnet 5.5 or Opus 5.5	Opus asks tougher, more strategic questions
Message house and narrative strategy	Opus 5.5	Strategic reasoning, nuance
Crisis holding statement, stakeholder Q&A, scenarios	Opus 5.5	High stakes; must anticipate second-order reactions
Monthly / quarterly client coverage report	Opus 5.5	Turns raw data into insight and finished documents
Award entries	Opus 5.5	Persuasive structure, evidence-led
New business proposals and RFP responses	Opus 5.5	Long briefs, competitive positioning
Annual reputation audit across thousands of articles	Fable 5.1	Long, multi-step research that must stay coherent

A day in a PR agency with Claude

Here’s how a PR professional can weave Claude through a normal working day without it becoming another tab you forget about.

8:00 · Morning briefing Haiku 5.5

Paste in the overnight coverage export (or let Claude search the web for news on your client and competitors) and ask for a one-page briefing: top stories, sentiment, which key messages landed, and anything that needs a response today. What took 45 minutes now takes five, plus your review.

9:30 · Media research and pitching Sonnet 5.5

Give Claude your story angle and a target journalist’s recent articles. Ask it why this reporter would care, what angle fits their beat, and to draft a short, personal pitch. You edit, check details and send from your own inbox. The relationship stays yours.

11:00 · Client call prep Sonnet 5.5

Upload the client brief, last month’s report and recent coverage. Ask for a one-page agenda, three proactive recommendations and the questions the client is likely to ask. After the call, paste the transcript and get action items, owners and a follow-up email draft.

13:00 · Press release drafting Sonnet 5.5 Opus 5.5 review

Draft the release with Sonnet using your client Project (brand voice, boilerplate, approved spokesperson names). Then ask Opus to review it as a skeptical business reporter: what’s the news, what’s missing, what would make them ignore it?

15:00 · Issues monitoring and crisis readiness Opus 5.5

When an issue starts bubbling, use Opus to map stakeholders, draft a holding statement and build a reactive Q&A, so legal and leadership review a strong draft rather than a blank page.

16:30 · Reporting and new business Opus 5.5

Turn a spreadsheet of coverage into a client-ready report with insights, charts and recommendations. Claude can create the document, slide deck or spreadsheet itself. For a big pitch, give Opus the RFP and your case studies and ask for a proposal outline and a competitive positioning section.

How to set up Claude for a PR agency

Most disappointing AI output comes from missing context, not a weak model. Ten minutes of setup per client fixes most of it.

1. Create one Claude Project per client

Projects are dedicated workspaces with their own files and instructions. For each client, add:

• Brand guidelines, tone of voice and style rules (AP style, banned words, spelling variant)
• Approved boilerplate, key messages and proof points
• Spokesperson names, titles and bios
• Your three best past press releases (as style examples)
• Current campaign brief, priority media and topics to avoid

Then write short project instructions, for example: “You are a senior PR writer for {{Client}}. Always follow the attached style guide, use only approved spokesperson quotes, and flag any claim that needs a source.”

2. Choose the right plan for an agency

Individuals can start on Pro. Agencies usually do better on the Team plan, which adds shared projects, central admin controls and organization-level connectors, or Enterprise when clients demand stricter security and data terms. Plans change often, so confirm details at claude.com/pricing.

3. Connect the tools your team already uses

Connectors let Claude read from and work with apps such as Google Drive, Gmail, Slack and Microsoft 365, so you can ask it to pull last quarter’s report from Drive or summarize a client email thread without copy-pasting. On Team and Enterprise plans, an admin enables connectors for the organization first.

4. Bring Claude to where PR work actually happens

• Claude for Microsoft 365 works inside Office apps, which is useful if your agency lives in Word, Outlook and PowerPoint.
• Claude in Chrome can act in your browser, which helps with research tasks across many web pages.
• Claude Tag brings a team-wide Claude agent into Slack, so account teams can ask for help right in client channels.

5. Turn your agency playbook into reusable skills

Skills are packaged instructions Claude follows for a specific kind of task. Package your press release template, pitch checklist or report structure as a skill, and every draft follows your house standard without anyone pasting the same instructions every time.

10 copy-ready Claude prompts for PR

Replace the bracketed parts, attach the relevant files, and run them inside the right client Project. The model tag shows the recommended starting point.

1. Press release from a brief Sonnet 5.5

Using the attached brief and our client's style guide, write a press release announcing [news].
Requirements:
- Headline under 12 words with the news, not adjectives
- Lead paragraph answers who, what, when, where, why in under 40 words
- One quote from [spokesperson name, title], marked [DRAFT QUOTE – NEEDS APPROVAL]
- Include the approved boilerplate unchanged
- Flag every statistic or claim that needs a source with [SOURCE?]
Then give me 3 alternative headlines and a 2-line explanation of what makes this newsworthy.

2. Personalized journalist pitch Sonnet 5.5

Here are [journalist name]'s last 5 articles at [outlet] and our story angle: [angle].
1. In 3 bullets, explain what this reporter consistently covers and what they would find interesting about our story.
2. Write a pitch email under 150 words: subject line, a first line that references their work specifically, the news, why it matters to their readers, and one clear offer (interview, data, exclusive).
3. Tell me honestly if this is a weak fit and who at the outlet might be better.

3. Media list clean-up and tiering Haiku 5.5

Attached is a media list. Remove duplicates, standardize outlet names, and add a "Tier" column:
Tier 1 = national/top trade outlets covering [topic], Tier 2 = regional or secondary trade, Tier 3 = everything else.
Add a "Fit note" column (max 12 words) explaining relevance to [campaign].
Do not invent emails or contact details. Mark any row with missing or suspicious data as "CHECK".
Return as a table.

4. Message house Opus 5.5

Build a message house for [client/initiative] using the attached materials.
Include: an umbrella message (1 sentence), 3 pillar messages, 3 proof points per pillar (only from the attached sources, cite which file), and audience variations for [audiences, e.g. investors, customers, employees, regulators].
Then list the 5 toughest questions a skeptical journalist would ask and the message gap each one exposes.

5. Spokesperson interview prep Opus 5.5

Act as [journalist name or type, e.g. a tough business reporter at a national daily] interviewing [spokesperson] about [topic].
First give me a one-page briefing: the reporter's likely angle, 3 key messages to land, and red lines to avoid.
Then run a mock interview: ask one question at a time, wait for my answer, and after each answer score it 1–5 for clarity, message discipline and quotability, with a suggested stronger version.

6. Coverage report with insights Opus 5.5

Attached is our coverage export for [client] for [period] and the campaign objectives.
Create a client-ready report:
- Executive summary (5 bullets, results first)
- Coverage by outlet tier, sentiment and key message pull-through
- Top 5 pieces and why they mattered
- Share of voice vs [competitors] if data allows
- What worked, what didn't, and 3 recommendations for next month
Show your calculations in a table so I can verify every number.

7. Red-team an announcement Opus 5.5

Read this announcement as three people: a skeptical trade reporter, a competitor's comms lead, and an angry customer on social media.
For each, tell me the first thing they would criticize, the headline they would write, and how we should adjust the announcement now to pre-empt it.

8. Thought leadership byline Sonnet 5.5

Draft a 700-word bylined article for [spokesperson] for [outlet], based on the attached interview transcript.
Use only opinions and facts the spokesperson actually expressed in the transcript. Keep their natural phrasing where possible.
Structure: a sharp, contrarian opening, 3 arguments with one concrete example each, and a forward-looking close. No promotional language about [company].
Suggest 3 titles.

9. Award entry Opus 5.5

Using the attached campaign results and the award criteria for [award/category], write an entry within [word limit].
Map every section directly to the judging criteria, lead with measurable results, and keep the narrative in the challenge → insight → strategy → execution → results structure.
List any criterion where our evidence is weak and what data would strengthen it.

10. New business proposal Opus 5.5 Fable 5.1 for large RFPs

Attached are the RFP from [prospect], our credentials deck and 3 relevant case studies.
1. Summarize what the prospect really needs (including what they don't say explicitly).
2. Propose a strategic idea with a one-line big idea, 3 workstreams and a 90-day plan.
3. Draft the "Why us" section using only our real case studies.
4. List the risks and assumptions we should clarify before the pitch meeting.

Using Claude for crisis communications

Crisis work is where AI for public relations pays off fastest, and where the guardrails matter most. In the first hour, the bottleneck is usually getting a solid first draft in front of leadership and legal. Claude, ideally Opus 5.5, can help you:

• Draft a holding statement in minutes that acknowledges the issue, shows concern and commits to updates, without admitting facts you haven’t confirmed.
• Map stakeholders (employees, customers, regulators, investors, partners, local communities) and tailor messages for each.
• Build a reactive Q&A covering the questions you hope no one asks.
• Run scenarios: “If this escalates to X, what changes in our response?”
• Check tone: ask Claude to flag anything that sounds defensive, legalistic or tone-deaf.

Two rules for crisis use: everything goes through legal and leadership approval, and you only feed Claude information your agency’s data policy allows on your plan. Better still, build crisis templates and scenario plans in Claude before you need them.

Claude, AI search and the new earned-media reality

Journalists, buyers and investors increasingly ask AI assistants questions instead of scrolling search results, and Muck Rack’s 2026 State of PR research put Generative Engine Optimization (GEO) on the PR agenda. That changes what “good coverage” means: clear, factual, well-sourced content is more likely to be understood and reused by AI systems.

You can use Claude to support this work by:

• Auditing a client’s newsroom and owned content for clarity, consistency and outdated facts
• Turning announcements into crisp fact sheets and FAQ pages
• Checking that key messages, product names and executive titles are consistent across every asset

Strong earned media in credible outlets remains the foundation. Claude helps you make the source material easier to understand and harder to misquote.

Guardrails: accuracy, confidentiality and ethics

Claude is powerful, but PR runs on trust. Build these rules into your agency AI policy:

1. Verify every fact, name, title, date and figure. Claude can be wrong. Ask it to flag unsupported claims and to show its sources or calculations.
2. Never fabricate quotes. AI-drafted quotes are placeholders until the spokesperson approves them, and they should be clearly marked as drafts.
3. Protect client data. Use business plans with admin controls, respect NDAs and embargoes, and keep legally privileged material out unless your policy explicitly allows it.
4. Keep journalist relationships human. Claude can research and draft; you decide who to pitch, and you hit send.
5. Be clear about disclosure. Agree with each client how and when AI-assisted work is disclosed.
6. Protect junior talent. Use Claude as a coach (ask it to critique drafts and explain its reasoning) so early-career staff still learn the craft.

FAQ: using Claude for PR

Is Claude good for PR work?

Yes. Claude is strong at the parts of PR that involve reading and writing at volume: drafting press releases and pitches, summarizing coverage, building message houses, preparing spokespeople and turning messy data into client reports. It works best as a fast first-draft partner and critic, with a PR professional checking facts, tone and judgment before anything goes out. Which Claude model is best for writing press releases?

Claude Sonnet 5.5 is the best default for press releases, pitches and everyday client copy because it balances quality and speed. Use Claude Opus 5.5 for high-stakes announcements, crisis statements or anything that must survive legal and executive review. Can Claude build a media list?

Claude can clean, de-duplicate, tier and annotate a media list you provide, and with web search it can research a journalist’s recent public coverage. It is not a verified contact database, so keep a dedicated media database as your source of truth and confirm every contact before outreach. Is Claude free for PR professionals?

Claude has a free plan, but the strongest models such as Claude Opus 5.5 require a paid plan. Agencies usually move to the Team plan for shared client projects and admin controls, or Enterprise for stricter data requirements. Check claude.com/pricing for current plans. Is it safe to put client information into Claude?

Treat it like any other vendor that touches client data. Use a business plan (Team or Enterprise) with admin controls, follow your client contracts and NDAs, avoid pasting embargoed or legally privileged material unless your policy allows it, and make sure your agency has a written AI-use policy. What is the difference between Claude Opus 5.5 and Claude Fable 5.1?

Claude Fable 5.1 is Anthropic’s Mythos-class model and its most capable publicly available tier, built for the longest and hardest multi-step work. Claude Opus 5.5 performs at a similar level on most tasks while being faster and cheaper, so it is the better everyday choice for PR strategy. Reserve Fable 5.1 for large research or reporting projects where Opus still falls short. Will Claude replace PR professionals?

No. Claude speeds up research and drafting, but PR runs on relationships, judgment and trust. Industry research shows almost all PR pros still edit AI output before using it, and only a small minority let AI agents act on their behalf. The professionals who gain the most use Claude to free time for strategy and journalist relationships.

The bottom line

The best way to use Claude for PR is to match the model to the job: Haiku 5.5 for volume, Sonnet 5.5 for everyday writing, Opus 5.5 for strategy and high-stakes work, and Fable 5.1 for the rare project that needs everything. Give it context through client Projects, connect it to the tools your team already uses, and keep humans in charge of facts, quotes and relationships. Do that, and Claude stops being a novelty and becomes the account team member who never runs out of time.

Sources: Muck Rack, State of AI in PR 2026 · The Batch: Claude Opus 5.5 Leaps Forward · The Batch: Claude Fills Out Its Model Family · Claude Platform docs: Claude Fable 5.1 · Anthropic: Project Glasswing · Claude Help Center: Connectors