Can AI make a media list?

The marking schemeLink to this section: The marking scheme§

How the media list was marked.

Carly builds the list for a real client brief by hand, the way she has for fourteen years. Hers is the answer key. Three things are measured against it, each multiplied by a public weight, and added up.

Weights add to 1.00 · no hidden multiplier · a name Carly says she’d include counts on top of her list, so a component — and the score — can pass 100 in theory

FoundDid we find them?

The share of the 23 journalists Carly pitched by hand that appeared on our list — plus the 11 she didn't have and said to add.The green +11 are finds of ours that weren't on her list and she said yes to when she graded our list. Her list stays what we're graded against, so a find she adopts lifts the score, even past 100.

Emails rightWould it even land?

Of the 12 of her journalists we found, the share with a working address. This time that counts an address at another domain of the same newsroom: CNBC's people take mail at more than one company domain, and Carly says each of them is live.

Good fitsWhen we say strong fit, is it?

Of the 81 names the tool stamped its strongest fits, the share on her list, standing next to one of hers at the same outlet, or a name she said yes to.

The pile so farLink to this section: The pile so far§

Every paper. Including the missing one.

One real client brief every two weeks (every week through Oct 1), with the mark — or the absence of one — left on the record.

1. 

   Week elevenOct 29, 2026

   Score skipped this week.

   Carly Martinetti is now Medialyst’s co-founder.

   After being a solo founder for 10 months, I now have a co-founder.

   Together we're building the PR operating system for the AI era.

   When Carly and I first started working together, she graded our AI 27.9 out of 100.

   Her words: "if we had pitched those reporters, our clients would have fired us."

   Embarrassing. Also exactly what the company needed. (tbh I had no idea how many times I actually had to rebuild the whole product to fix that.)

   But that's when I knew the right people were in the room. Carly knows the subtle things that make a reporter actually write the story. And I can turn all of that into software by Thursday. If we can't crack PR for the AI era together, I'm not sure who will.

   So we built. And Medialyst got better every week.

   Along the way, some customers started treating her like a co-founder. I treated her like a co-founder, and she was already acting like a co-founder: protecting the bar in every corner of the product.

   Since that call in July, Medialyst has grown 3.4x, almost entirely organically. Most of what moved it came out of her red pen.

   Working together showed us a glimpse of what AI+PR actually should be:

   context → opportunity → story → journalist → relationship → coverage → AI visibility → learning → context

   Watching this loop run inside Notably convinced us PR needs more than an AI media list builder. It needs an entirely new operating system, built for the AI era.

   The solo-founder version of Medialyst was peanut butter: a cute tool people occasionally found useful.

   The full sandwich is what PR runs on.

   Welcome to the sandwich, Carly.

   Read it on LinkedIn ↗
2. New brief, her new keyOct 15, 2026Link to this section: New brief, her new key, Oct 15, 2026§−0.9 vs last run, on a new brief and key

   The desk, not just the byline

   Graded on: A personal-finance story that landed in Fortune

   Last time was breaking news: Carly was rushed and the story was big, which favored the AI. This time was a personal-finance story that landed in Fortune, and she built her list by hand: 23 journalists. A harder, truer test. Medialyst found 12 of her 23, plus 11 she didn’t have and said yes to.

   • 12+11/23found▲ 5.2pts · was 37/58
   • 12/12emails right▲ 8.1pts · was 34/37
   • 30/81good fits▼ 18.0pts · was 82/149

   B

   81.1/100

   What we changed to get here

   • Outlet Sweep now finds who covers your story’s desk at each outlet, even if they haven’t written this exact story yet.
   • Outlet Sweep also adds the editor who runs that desk, so you can pitch them when there’s no obvious reporter.
   • Top-tier pitch angles. When your pitch is missing something — a WSJ or NYT story usually needs a real person in it — the pitch angle now tells you what to add to make it land.
   • Free journalist lookup. Anyone can now look up a journalist for free with our Journalist Finder. All those lookups surfaced errors in our journalist database fast, and we fixed them at the source: journalists who changed outlets now show where they work today, rows carry the outlet we have evidence for, work addresses rank ahead of personal ones, and anyone can report wrong info in one click.

   MarkedElvis SunCo-founder, Medialyst

3. New brief, her new keyOct 1, 2026Link to this section: New brief, her new key, Oct 1, 2026§+10.9 vs last run, on a new brief and key

   Outlet by outlet, including the ones AI can’t read

   Graded on: A quick-response pitch about Trump TV

   Carly had to jump on two breaking news stories for separate clients this week, so she used Medialyst to source contacts for one of them. It found several journalists she wouldn’t have thought of — and they turned into actual stories for her client. A new brief, so a new key: the 58 journalists she’d have pitched by hand. The score grades the list Medialyst built against them, plus the names she said yes to.

   • 37+18/58found▲ 34.5pts · was 36/63
   • 34/37emails right▲ 5.8pts · was 31/36
   • 82/149good fits▼ 15.4pts · was 31/44

   B

   82.0/100

   What we changed to get here

   • Outlet Sweep. Give Medialyst your story and it now picks the outlets that matter most for it. Then it finds the right person at each one: who writes there now, who covers your angle, and who’s the best fit. It went live this week and hasn’t added any of Carly’s names yet.
   • Reuters, too. Reuters blocks AI from reading its stories, so we haven’t been surfacing anyone from Reuters, even when it was the right outlet for your story. Now, with Outlet Sweep, we finally can. The same goes for the FT, The Times and The Economist.
   • Project memory works. For this story, Carly taught Medialyst the difference between a White House reporter and a legal correspondent, and it picked that up right away.

   • The full story, on our blog ↗

   MarkedElvis SunCo-founder, Medialyst

4. Same list, deeper readSep 24, 2026Link to this section: Same list, deeper read, Sep 24, 2026§+3.9 vs last run, same brief, key and list

   We read deeper. The score went up.

   Graded on: A small-business survey brief

   No new list this week. We re-scored last week’s and changed only how journalists are scored. Fewer names now come back as strong fits, and the reporters Carly would only send an FYI land where they belong, in the 60s to 80s.

   “I C what you did there… You improved medialyst!!”

   • 36+2/63found— no change · was 36/63
   • 31/36emails right— no change · was 31/36
   • 31/44good fits▲ 12.9pts · was 42/73

   C

   71.1/100

   What changed

   • From one article to their whole byline. We used to score a journalist on the one article the row was found on. Now we read their author page, their recent bylines and the openings of their latest articles first.
   • Who covers your topic vs. who will actually write your story. The Wall Street Journal’s editor-in-chief covers the economy, but they are not writing up the survey you paid for, so they no longer land in your strong band.
   • Carly’s instincts, written down. She graded our lists by hand and told us why each name was out. Those reasons are rules now: paid placements, guest essayists, as-told-to first-person desks, contributor networks where a pitch goes nowhere, brand-content pages, newswires — a dozen signals someone who pitches every day spots in seconds. Fun fact: we asked for feedback and she sent 1.8 GB of it, on Loom.
   • Pro and Scale first. The deeper read is live on Pro and Scale plans today. Other plans follow once we have watched quality, cost and the load it puts on the rest of the system.
   • Also shipped: Radar. One per project. Every morning it reads the news against what each client can credibly stand behind and sorts the stories into pitch now, big story and watch, with the angles and the evidence behind each one.

   • The full story, on our blog ↗

   MarkedElvis SunCo-founder, Medialyst

5. Her folder in the roomSep 17, 2026Link to this section: Her folder in the room, Sep 17, 2026§+6.8 vs last run, same brief and key

   The agent started where Carly starts

   Graded on: A small-business survey brief

   Same brief and the same 63-name key as last week, so the arrow is like for like. What changed is what the agent knew before it started. Last week it had the chat box. This week it had a project: the client’s folder, the onboarding doc with the dream publications, the full campaign brief, and a page of Carly’s rules in her words.

   • 36+2/63found▲ 7.9pts · was 31/63
   • 31/36emails right▼ 1.0pts · was 27/31
   • 42/73good fits▲ 13.1pts · was 40/90

   D

   67.2/100

   What we changed to get here

   • Projects. A project is a client, a brand, or one story. Everything you give the agent about that client lives there, and every campaign picks one. Her list was built inside a project for the first time.
   • Project files. The brief, the press release, the onboarding doc, last year’s list, the coverage tracker. The agent reads the relevant ones before it proposes an angle, and scoring reads a digest of what’s relevant in the project.
   • Agent memory. One plain-text file per project, written three ways: typed in, proposed from a “this score is wrong” click, or noted from the chat.
   • Receipt: one memory line named the outlets she cares about. The brief named none. The plan came back with 11 outlet-pinned searches, and those searches produced 12 journalists on her list.
   • Also this week: Beat Sweep Agents got a nudge to continue until they hit their sweep budget. It now follows the list size you asked for more closely.

   • The full story, on our blog ↗

   MarkedElvis SunCo-founder, Medialyst

6. New brief, her new keySep 10, 2026Link to this section: New brief, her new key, Sep 10, 2026§−5.9 vs last run, on a new brief and key

   A quiet week: the addresses held, the list didn’t move

   Graded on: A small-business survey brief, under embargo

   Light week. Carly built this list herself in the app for a real pitch, so for the first time the graded list is hers end to end — no brief from us. She keyed 63 journalists. We found 31 of them, plus two more she adopted. She checked every address we returned: 27 of 31 right.

   • 31+2/63found▼ 14.3pts · was 7/12
   • 27/31emails right▼ 12.9pts · was 7/7
   • 40/90good fits▲ 12.3pts · was 9/28

   D

   60.4/100

   What changed, and what didn’t

   • The database, not the product. Most of the month went into the media database: addresses promoted from what we already hold, late finds written back, non-person bylines kept away from paid lookups. Last week’s 7 for 7 on emails got lucky; 27 of 31 is closer to what the database actually is.
   • Two of her four missed addresses were ours all along — one sat on a duplicate record of the same freelancer, one attached to his record 34 seconds after her row finished. Both are fixable without buying anything.
   • The feature we planned to ship this week got pulled at the last minute. Carly graded it side by side against the current version and didn’t think it was good enough. Back to work until it is.
   • Receipt: the first time a WSJ byline has appeared on any of her goldens — three of hers this week, from a new infrastructure upgrade we put in this week.

   • Elvis’s fixes, on LinkedIn ↗

   MarkedElvis SunCo-founder, Medialyst

7. Week fiveSep 3, 2026Link to this section: Week five, Sep 3, 2026§no new mark

   The score is absent. The week wasn’t.

   Recorded as: Absent

   We didn’t complete the weekly grading by Thursday, so there is no new Carly Score to publish.

   Absent

   No score: —/100

   Latest verified66.3 · Aug 27

   Where the week went

   Our free Claude Routine for journalist request scanning went viral. We spent most of the week on system stability and customer support for existing users. No new changes were made to the core system.

   • The full story, on LinkedIn ↗

   AbsentElvis SunCo-founder, Medialyst

8. New brief, her new keyAug 27, 2026Link to this section: New brief, her new key, Aug 27, 2026§+15.0 vs last run, on a new brief and key

   First mark off the F band — and one find she adopted

   Graded on: A niche NFL announcement brief

   A new brief this week — a real client announcement that has not gone out yet, so the card describes it rather than names it. Carly built a fresh 12-name key for it and graded every row herself: 7 of her 12 found, every served address right, and 9 keeps from the 28 names the tool vouched for hardest. One keep was not on her key at all — a find of ours she said she'd include, and under the bonus rule that counts on top of her list, in green.

   “D stands for… doing great?”

   • 7+1/12found▲ 27.1pts · was 17/43
   • 7/7emails right▲ 11.8pts · was 15/17
   • 9/28good fits▲ 2.0pts · was 25/83

   D

   66.3/100

   What we changed to get here

   • Angle control. The plan proposes the story angles and you rule on them — how many run, which ones, in what order — before anything spends a credit. The angles are yours to cut, keep, or add to.
   • Research depth. How long each Beat Sweep Agent digs is a dial now, not a default — rounds per angle, and how far afield it reads before it stops.
   • Spend control. Every run carries a budget envelope, per angle and in total, and the research stops at the cap instead of the bill saying what happened.
   • Receipt: the journalist she adopted — the green +1 — came out of a comparable-story angle the planner proposed itself and the run approved. That lane exists because angles are explicit now, ruled on before anything runs.
   • Receipt: five angles deep, the run took about six minutes of agent research from plan approval to her second anchor article — inside its spend envelope, with budget to spare.

   • Carly’s grading, on LinkedIn ↗
   • Elvis’s fixes, on LinkedIn ↗

   MarkedElvis SunCo-founder, Medialyst

9. Same brief, rerunAug 20, 2026Link to this section: Same brief, rerun, Aug 20, 2026§+9.1 vs last run

   From 7 of her 43 journalists to 17, in one week

   Graded on: A job site’s Stress-Free Six-Figure Careers brief

   Finding her journalists is 40% of the score and it had not moved in a month: 7 of her 43, three gradings running. This week it moved to 17. The list was committed and dated before anything was compared to her sheet, and the score is the same arithmetic every mark on this page gets.

   • 17/43found▲ 23.3pts · was 7/43
   • 15/17emails right▲ 2.5pts · was 6/7
   • 25/83good fits▼ 3.2pts · was 7/21

   F

   51.3/100

   What we changed to get here

   • Angles. Your news is never one story. The plan now breaks it into several distinct story angles — each written the way the reporter who covers that beat would want to hear it, each going after its own set of journalists — and you see and steer the angles before anything runs.
   • Beat Sweep Agents. One research agent takes each angle and works it the way a person would: reading what it finds, deciding what to look up next, hunting the rival reports yours is up against and who covered them by name, and checking who actually reads an outlet before vouching for anyone.
   • Receipt: two of her journalists whose exact articles sat unread on our own search pages last run are on the list this time — reading the whole pile caught exactly the misses the audit predicted.
   • Receipt: the rival-report hunt lands people topic searches never surfaced — several of the ten new finds came in through competing rankings they had covered, at outlets that had never appeared in any earlier run of this brief.
   • Receipt: off-target outlets among the names it vouched for halved, and 15 of the 17 addresses were right — wider search did not come at the cost of a messier list.

   • The full story, on our blog ↗

   MarkedElvis SunCo-founder, Medialyst

10. Same brief, rerunAug 13, 2026Link to this section: Same brief, rerun, Aug 13, 2026§+8.5 vs last run

    We fixed the addresses everywhere, not on her sheet

    Graded on: A job site’s Stress-Free Six-Figure Careers brief

    Only the addresses moved — finding her journalists carries the most weight, it did not budge, and that is still where the score is lost. One address stays a miss we are disputing until she rules.

    • 7/43found— no change · was 7/43
    • 6/7emails right▲ 28.6pts · was 4/7
    • 7/21good fits— no change · was 7/21

    F

    42.2/100

    What moved underneath this mark

    • More of your lists come back with an email. Last Friday, 53.5% of requests came back with a verified, deliverable address. Today it is 68.3% — up 14.9 points. We think the theoretical ceiling is somewhere between 80% and 90%, because not every news article published has a journalist with an email you can reach.
    • Freelancers stopped being the worst group on the list. Last Friday we could reach 39.4% of the freelancers our users asked for, and about 1 in 8 of the journalists our users ask about is a freelancer. We put a deep research agent on them — it reads the open web, bylines and social profiles to find the address. They are now at 62.2%, up 22.8 points.
    • Lists build faster — a usable list in 2.6 minutes. Half your list comes back in about 2.6 minutes on a typical build; in January that was 8.3 minutes. The slowest lists used to run into the 60-minute timeout, and now finish in 6.6.
    • The reason it got faster: the database learned. In February almost every email had to be hunted live while you waited. In our measured re-run this week, 86.6% of the addresses we served came straight from the database, the rest found in real time by our deep research agents — and we did not need a single lookup from a third-party contact provider.
    • We worked out who the top 1,000 journalists are, and now we grade ourselves on them. Of the top 1,000 most-requested journalists, 95.3% now have a verified email. Of the top 10,000, 84.5% do.

    • The full story, on our blog ↗

    MarkedElvis SunCo-founder, Medialyst

11. Same brief, rerunAug 6, 2026Link to this section: Same brief, rerun, Aug 6, 2026§+5.8 vs last run

    We stopped vouching for journalists we could not verify

    Graded on: A job site’s Stress-Free Six-Figure Careers brief

    The same brief run fresh through the product two weeks later. What moved: more of the names it vouched for are ones she would actually pitch.

    • 7/43found— no change · was 7/43
    • 4/7emails right— no change · was 4/7
    • 7/21good fits▲ 19.3pts · was 7/50

    F

    33.7/100

    What we changed to get here

    • When we could barely find a journalist's published work, we now go looking for more of it before judging whether they fit.
    • When we still cannot find enough of it, we say “not enough evidence” instead of inventing a confident-looking score.
    • An author's index page no longer counts as an article they wrote.
    • An email or LinkedIn profile that matches a completely different name is no longer treated as a confirmed match.
    • Articles the assistant adds to an imported list now actually get turned into journalists.

    • Carly’s grading, on LinkedIn ↗
    • Elvis’s fixes, on LinkedIn ↗

    MarkedElvis SunCo-founder, Medialyst

12. BaselineJul 20, 2026Link to this section: Baseline, Jul 20, 2026§first mark

    The floor, before a single fix

    Graded on: A job site’s Stress-Free Six-Figure Careers brief

    The list our product built, walked through with her line by line.

    “Not yet, if we pitched these journalists our client would fire us.”

    • 7/43found
    • 4/7emails right
    • 7/50good fits

    F

    27.9/100

    Why this is the first mark

    Seven months building this, with nothing to grade it against — we genuinely did not know whether the lists were any good. Then Carly joined Medialyst and marked the first one we handed her. She still does not go easy on us. Now she is our co-founder, and the Carly Score keeps running.

    • Carly’s note, on LinkedIn ↗
    • Elvis’s note, on LinkedIn ↗

    MarkedElvis SunCo-founder, Medialyst

Your turnLink to this section: Grade our list§

Want to grade our list?

Run your own brief. Then do the thing you would do anyway before you hit send: go down the list and cut the names you would cut. Tell us how many survived and what we got wrong.

Post it wherever you post, send us the link, and we will pin it here. We do not edit it.

• Pros Very campaign specific Easy to use Cons - Not a replacement for MuckRack yet - Didn't see a way to compare the list that you compiled based off my pitch compared to a list I already currently had

• Apparently, telling the AI “make no mistake” does not improve the quality of the list.

• A lot of these are advertorial articles. Can you pick up earned coverage? I want to see more broadly stories from the restaurant and chef business world. These focus too much on delivery.

• can you resort the list to make it only north america? I see indonesia. And many of these emails are not deliverabke

• i need top tier national outlets, not regional

• this media list is only supposed to be local albany publications and reporters

• you are surfacing a lot of national journalists based in NYC. I only want to see people from New York specific media outlets who cover NYC.

• I do not think we want people who write about Indianapolis already; that will result in local writers. We want national sports, economics, or leadership writers

Questions and answersLink to this section: Questions and answers§

Questions about the score.

What is the Carly Score™?

The Carly Score™ is a public mark on Medialyst's media lists, published every two weeks. Carly Martinetti, a publicist who has built media lists by hand for fourteen years, builds a list for a real client brief. Medialyst is given the same brief. Her list is the answer key, and ours is marked against it out of 100. We aim to publish the number every other Thursday, whatever it says. If a mark is not completed, the missing mark stays on this page too.

Why is the Carly Score™ so low?

Because the test is deliberately hard. The score marks a one-shot run: Medialyst gets the brief — Carly's word for the client's description of the story and who should cover it — in a single prompt, and has to reproduce, unassisted, the exact list a senior publicist spends days building by hand. No follow-up prompts, no steering, no second pass. It is the strictest marking we know how to set.

Does a Carly Score™ of 81.1 mean Medialyst is 81% accurate?

No. The score measures one thing: whether an unassisted agent can one-shot, from a single prompt, the media list a senior publicist like Carly builds over a few days. The number measures how much additional work a list needs before it matches hers — not accuracy.

Can users get a better media list than the Carly Score™ suggests?

Yes. The score grades the raw first draft. The product is also designed for what happens next: you and the agent iterating on the list together — expand it, steer it toward a beat, cut the names that do not survive a closer look, fix an email by hand. That iteration is deliberately excluded from the graded run. A Medialyst list is a starting point, never a finish line — the Carly Score™ measures the starting point. The first list is free, so run your own brief and mark us yourself.

How is the Carly Score™ calculated?

Three measures, each multiplied by a published weight: the share of her journalists we found (weight 0.40), the share of those whose email address matched hers (0.30), and the share of the names we stamped a strong fit that she would actually pitch (0.30). The weights add to 1.00 and there is no hidden multiplier. Every published mark, old or new, is scored under the same formula.

Is it a new brief every time?

We try to — a brief that repeats is a test you can accidentally tune for, and rotating briefs is the guard against that. When a brief does repeat, its report card says so and calls the mark a rerun. A rerun is still a real measurement, though. Medialyst finds journalists through what they are writing right now, so running the same brief at a different time produces a different list by design. And journalists move beats and change outlets, so each graded run is marked fresh as Carly works the list again, not against a frozen answer key.

How does the Carly Score™ compare Medialyst to Muck Rack, Cision, or Meltwater?

It does not. No competitor has ever asked Carly to grade one of their lists, so we cannot give you their number — not out of politeness, we just don't have one. Any of them is welcome to fix that: email [email protected] and we will publish their Carly Score™ on this page, logo attached, whatever it says. The feature-by-feature comparison lives at medialyst.ai/compare.

Who grades the Carly Score™?

Carly Martinetti, co-founder of Medialyst and the PR agency Notably. She has been building media lists by hand for fourteen years. She marks the list independently, and we publish her number without editing it.

How often is the Carly Score™ updated, and what is it now?

Every two weeks, on a Thursday — the next mark is due Oct 29, 2026. It came out every week through Oct 1, 2026. Each mark is dated on this page along with the counts behind it, and a missed mark stays on the record instead of becoming a zero or a repeated score. The most recent verified mark is 81.1 out of 100, graded Oct 15, 2026. No new mark was issued Sep 3, 2026 or Oct 29, 2026.