AI Media List Tools: Muck Rack vs Semrush AI PR Toolkit vs ChatGPT

Muck Rack is the safest base for an agency that pitches every week, because it is built on a journalist database. The Semrush AI PR Toolkit suits teams wanting a published price (from $149 a month, billed annually). ChatGPT is a strong research assistant but cannot verify contacts. Test all three on one list before committing.

Rankings of these tools mostly repeat feature lists. The more useful comparison takes an afternoon: build one media list three ways from the same pitch, check every name against a source, and work out the cost per verified journalist. Vendor claims about accuracy, including Muck Rack's own beta figures, are no substitute for that number.

Follow me on Instagram@sagnikteaches

What each of the three routes gives a PR team

Muck Rack is a PR platform built on a journalist database. It launched its Media List Agent in beta on 2 February 2026. You describe the journalists you want and the agent returns up to 50 suggestions, each with an explanation of the match and the relevant articles; you refine them in plain language, and while you write a pitch it can suggest journalists to add or remove. At launch it handled English-language pitch content only, and it recommends without sending anything, so approval stays with you. Muck Rack reports that beta customers saw an average 10 per cent rise in pitch open and click rates, a vendor figure rather than independent data. Its pricing page lists Starter, Standard and Premier plans with no prices, offers a customised proposal and limits media list exports on Starter, so ask which plan includes the agent; Muck Rack has since announced a broader Muck Rack Agent as well.

Connect on LinkedInSagnik Bhattacharya

Prowly is now the Semrush AI PR Toolkit; Prowly's own site says so. Semrush's PR pricing page lists three plans, billed annually: Base at $149 a month with 250 contacts from the media database each month, Pro at $279 with 500, and Business at $499 with 1,000, plus custom Enterprise pricing and extra users as an add-on. AI-driven journalist recommendations come with every plan, but exporting contacts starts at Pro, which matters if your verification happens in a spreadsheet. Review sites still quote older Prowly plans and prices, so go by the live page.

Subscribe on YouTube@sagnikteaches

ChatGPT is a general assistant. It has no journalist database of its own. With web search on, it can find recent articles and bylines, and it can classify, segment and rank a list you give it. Plus lists at $20 a month. Business plans do not train on your content by default, which matters when a pitch is under embargo.

Three jobs a media list has to do

A media list has three jobs: find the right people, hold accurate contact details, and tell you why each person fits this pitch. The three tools split those jobs differently. A database tool is strongest on the second, an AI agent on the first and third, and a general assistant on the third and on turning notes into segments. So the test must measure each job separately rather than ask which list "looks best".

The test: one pitch, three routes

Use a real pitch, not a made-up one. The example below is an illustrative brief for a marketing agency: a story about a software company's analysis of how small firms handle invoicing, built on data the company owns. Replace it with your own.

PITCH BRIEF (use identically in all three tools)
Story: A small-business accounting software company has analysed its own anonymised usage data on late invoices.
Hook: Invoices raised on a Friday are paid more slowly than those raised on a Tuesday, according to the company's data.
Spokesperson: the company's head of product.
Assets: data summary, three charts, a short quote.
Target: journalists who write about small-business finance, freelancing or accounting software.
Embargo: none.
Size of list wanted: 40 names.

Route 1: Muck Rack

Open the Media List Agent and paste the brief as the description of the journalists you want. Keep the explanation text the agent returns for each name. Refine once ("exclude broadcast, include newsletter writers") and record how many of the 40 changed. Export the list with every field the tool gives you.

Route 2: Semrush AI PR Toolkit

Use its media database search with the same keywords and filters (topic, outlet type, language), and its AI journalist recommendations, which every plan includes. Record which filters the tool supports for beats and topics, whether it shows recent articles per journalist, and how much of your plan's monthly contact allowance the test uses. Base does not export contacts, so either copy the verification fields by hand or ask Semrush to run the test on a Pro account.

Route 3: ChatGPT

Use a prompt that forces evidence, with web search on. The prompt below is an example.

You are a PR researcher. Using web search, find 40 journalists, columnists or newsletter writers who have written about small-business finance, invoicing, freelancing or accounting software in the last 12 months.
For each, give: name, outlet, role, one recent article title with its URL, why this pitch fits in one sentence, and a confidence score from 1 to 5.
Rules: if you cannot find a recent article, leave the name out. Do not guess email addresses. Do not include anyone you cannot link to an article.
Return a table.

Expect ChatGPT's table to be neat and partly wrong. That is normal, and the next step is why the test exists.

Verify every name: the part that decides the winner

Take all three lists, merge them, and check each name against sources you control. For each journalist, answer five questions in a spreadsheet: Is this person still at the outlet? Have they written on this topic in the last twelve months? Is the contact route correct? Is it the right kind of contact for a pitch (some staff do not take pitches)? Would a human on your team pitch them?

An illustrative results sheet for 40 names per tool:

These numbers are invented to show the method, not measured results; run your own. The shape is what matters. A database-backed tool is likely to score well on contacts, a chat assistant on speed and fit, and the cost of verification differs between them. Add total minutes and your hourly rate, then you have a cost per verified name, which is the number to compare.

For a worked sum: if the tool plus verification costs 45 minutes for 29 verified names, and your time is worth $60 an hour, that is $45 of time for 29 names, about $1.55 per name, before the subscription. If another route gives 21 verified names for 70 minutes, that is $70 for 21, about $3.33 per name. That, rather than the feature table, is the comparison.

Where the three lists overlap, and where they differ

Merge the three lists and count the overlap. Journalists named by all three are your core, and they deserve your best pitch. Names appearing in only one list need the closest checking, because they are either a clever find or an error. In an illustrative merge, 14 names appear in all three lists, 22 in two and 34 in one only (14 x 3 + 22 x 2 + 34 = 120 entries, 70 different people). Check the 34 first: you will usually find that ChatGPT's unique names are the riskiest and the database tool's unique names are the most reliable on contact details but may be weaker on topical fit.

Keep that merged sheet. It becomes your list, your audit trail, and the starting point for your next campaign.

What each tool is likely to get wrong

• Database tools. Out-of-date beats. A journalist who covered finance two years ago now covers politics, and the profile still says finance. Look at the last three bylines, not the tag.
• AI agents on top of a database. Over-confident match explanations. A line such as "covers small-business technology" may rest on a single article. Click through to the cited articles.
• ChatGPT. Invented or stale details: a plausible name at the wrong outlet, an article title that does not exist, or a link that returns an error. Insist on links and open every one.
• All of them. Contacts for people who do not accept pitches. Some columnists, editors and freelancers say so in their bios or on their outlet's contact page.

The step-by-step method for finding and segmenting is in media list building with AI; use it alongside this test.

From list to pitch: the same pitch through three routes

The test is not complete until you pitch. Take ten verified names from each route, split them evenly, and send the same pitch with the same subject line style, with one personal sentence per journalist referring to their article. Track opens, replies and any bounces over seven days. Ten names per route is too few to prove anything statistically, so use this only as a sanity check; the verification sheet is the stronger evidence. An illustrative personalised opening:

Hi [first name],

I read your piece on [article title] and the point about late payment made me think of some data a client has: invoices raised on a Friday are paid more slowly than those raised mid-week. Three charts and the head of product are available if useful.

Notice that the opener is only honest if the article exists and says that. This is why the verification step comes before the pitch. AI PR pitch emails with a review step covers the writing side.

Segmenting the merged list with ChatGPT

Once the sheet is verified, ChatGPT is at its best: sorting. Paste the verified rows (names, outlets, recent article titles) and ask for segments. An illustrative prompt and output:

Group these 36 verified journalists into three segments for the same pitch: (1) data-led finance reporters, (2) freelancer and small-business advice writers, (3) newsletter authors. For each segment, give a one-line angle and the one chart from the pitch that fits best. Do not add names.

A realistic reply gives you segment 1 with 12 names and the angle "the Friday effect, with the headline chart", segment 2 with 15 names and "what a freelancer should do on invoice day", and segment 3 with nine names and "the data in three sentences with a link". What you fix: check it has not dropped anyone, check it has not added a name (a common slip), and rewrite any angle that claims more than the data shows. The pitch about a Friday effect says what the company's data suggests; it does not prove why.

When a ChatGPT list pitched the wrong editor

An illustrative case. An agency takes a ChatGPT list at face value and pitches 40 names. Six bounce, four replies say "I don't cover this any more", and one editor replies, politely, that the pitch quoted an article she did not write. The article belonged to a colleague with a similar name. The agency's sender reputation takes a small dent, and the editor remembers the agency's name for the wrong reason. The fix is dull: open every linked article, confirm the byline, and confirm the person is still listed on the outlet's staff or contributor page. The quarter of an hour this takes per ten names is the cheapest insurance in the process.

Keeping a verified list fresh between campaigns

A verified list decays. Journalists move, outlets close and beats change. Keep one master sheet per client and add three columns: last verified date, last pitched date and outcome. Every month, re-verify anyone not checked in 90 days, and before every campaign re-check the top 20 names. Delete anyone who asks to be removed, the same day, and record it so no tool re-imports them. If a database tool is your source, do not assume its own freshness: check a sample of ten names each quarter and record how many were stale. That figure tells you whether the subscription is delivering the accuracy you are paying for, and it is a far better renewal argument than a sales call.

Questions to ask each vendor before you pay

1. How many contacts, searches and exports does my plan include each month?
2. How often are contact details re-verified, and how do I report a wrong one?
3. Does the AI show its evidence (recent articles) for each suggested journalist?
4. Are my pitch text and list data used to train models?
5. Can I export my lists if I leave, and in what format?
6. What is the contract length, and what notice is needed to cancel?

Write the answers into the results sheet next to the test scores, so price, accuracy and terms sit in one place. A tool that scores well but will not answer question 5 deserves a second thought.

Two short examples of lists tuned to the pitch

A narrow pitch: the data story

For the invoice-timing study, the right list is small. A tight list of 25 might hold eight data-led finance reporters, ten advice writers who regularly cite figures, and seven newsletter authors with a small-business readership. Every name has a recent article on late payment or invoicing. A broad business desk, however large its audience, is left off, because nobody there has written on the subject and the pitch would be deleted unread.

A broad pitch: a product launch

For a launch with no data hook, the list widens: trade titles for the sector, technology reporters who cover small-business software, and a few podcast hosts. Here the AI agent's explanation lines matter most, since they show why a name was suggested and let you drop weak matches quickly. In this case ChatGPT's role shifts to drafting three angles, one for each segment, and a human picks which to send.

The lesson in both is the same: the list is a function of the pitch, so a list built for one story should not be reused for the next without a fresh check. Build the habit of starting each campaign with the brief, running the verification sheet, and only then touching the send button.

How long the whole test takes

Allow about half a day: 30 minutes to write the brief, an hour to run all three routes, an hour and a half to verify the merged list, and an hour to pitch a small sample and tabulate. Put the figures in one table, record the cost per verified name for each tool, and make the decision at the end of the afternoon rather than after weeks of free trials. A shared sheet also lets a colleague repeat the test in six months, when the tools have changed and the answer might too.

When a database subscription pays for itself

Price the tools by the number of names you pitch, not by the headline subscription. Take a small agency running six campaigns a month with 40 names each: 240 names, which fits inside the Semrush Base plan's 250 database contacts a month. In the illustrative results above, the database route cost about $1.55 of staff time per verified name and the ChatGPT route about $2.88 (72 minutes for 25 names). Across 240 names that gap is about $320 a month, so a database tool costing up to roughly $340 a month, against ChatGPT Plus at $20, pays for itself on time alone; one costing more needs another reason, such as monitoring or reporting. Conversely, an agency running two campaigns a quarter will not recover a database tool's cost, and a ChatGPT-plus-manual-check routine is fine.

A rule of thumb to test, not to trust: if you pitch journalists weekly, a database tool earns its fee; if monthly, a cheaper plan or ChatGPT with disciplined checking might do; if rarely, do not subscribe, and ask a database vendor for a one-off trial instead. For each vendor, ask in writing for the price with your team size, the number of contact exports, the contract term and the cancellation terms.

Embargoed pitches and client confidentiality

A media list reveals a client's plans, and a pitch brief often contains embargoed news. Before pasting a brief into any tool, check where the data goes and whether it trains models. For ChatGPT, business plans do not train on content by default, and consumer plans have a setting to switch training off. For the other two, read the vendor's terms and ask in writing. Keep embargoed details out of the first search prompt when you can: describe the topic, not the news. The tutorial on what to check in an AI tool's privacy policy and terms lists the questions, and an approval process for new AI tools shows how to get a tool signed off inside the agency.

A starting point for weekly, monthly and occasional pitchers

• Agency pitching every week for several clients. A database-backed tool, with Muck Rack the one to put through the test first, and ChatGPT alongside for segmenting and draft angles.
• In-house PR person at a small company. Start with the Semrush AI PR Toolkit if you want a published price and no sales call, or ChatGPT with disciplined verification. Run the test on your next launch.
• Freelancer with a few clients. ChatGPT plus manual verification, with a one-month database trial only when a campaign needs wide coverage.

Whatever you choose, keep a human sign-off on each list, and track results after the send, which AI media monitoring for small PR agencies explains. For the general principles, the AI hub is the place to start.

If you would rather map this with someone, my AI implementation consultation covers that kind of tool and review-step decision; the details are on the contact page.

Questions about AI media lists

Is Prowly the same as the Semrush AI PR Toolkit?

Prowly now sits under Semrush and its pricing page is branded as the Semrush AI PR Toolkit, so older reviews of Prowly describe the same product family. Plan names, contact limits and prices have changed, so read the current page and ignore figures from older review sites.

Can ChatGPT build a media list on its own?

It can suggest outlets, beats and the kind of journalist to target, and it can search the web where search is on. It has no verified contact database, so names and emails must be checked against the outlet's own pages before anyone is pitched. Treat it as a research and segmenting assistant.

How many names should a first list have?

Fewer than you think. Thirty to fifty well-matched journalists beat several hundred loosely matched ones, because each pitch should reference something the person actually wrote. Muck Rack's Media List Agent itself returns up to 50 suggestions at a time.

What is the biggest risk with AI-built lists?

Pitching someone who changed beats, left the outlet or never covered the topic. It wastes goodwill and can hurt your sender reputation. A named check against the journalist's last three articles removes most of that risk.

Further reads

• AI Influencer Marketing for Agencies: Shortlist, Vet and Pitch — The same shortlist-and-vet method for creators.
• How Small Agencies Use AI to Write New-Business Pitches — Use AI on the agency's own new-business pitches.
• AI Content Workflow for Agencies: From Brief to Approved Draft — A brief-to-approved-draft workflow for agency content.
• Social Media Pre-Publish Checklist for AI-Drafted Posts — A pre-publish check for AI-drafted social posts.
• Marketing Agency Pricing After AI: Retainers, Packages and Deliverables — Price PR and content work once AI speeds it up.
• Crisis Communication Plan Template: Draft Holding Statements With AI — A filled-in crisis plan: scenarios, holding statement templates, an approval chain and a media Q&A drill, with AI drafting and named people approving.
• PR Coverage Report: Build One for Each Client With AI — A coverage report structure, the clippings log columns, AI prompts for the summary, and a sample monthly report with the checks a person must do.
• PR for AI Search: How Agencies Earn Citations in ChatGPT Answers — A citation test you can run this week, the coverage types worth pitching, a tracking sheet, client reporting wording and claims to avoid.
• AI Agents for PR: What They Do and How an Agency Supervises Them — A PR agency supervision method for media lists, monitoring and pitch drafts, with a strict approval-before-send gate.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Muck Rack's Media List Agent launch release (2 February 2026) and pricing page, Prowly's site, Semrush's AI PR Toolkit pricing page, OpenAI's ChatGPT pricing (read 1 November 2026; verify current plans and limits before buying).