Bersyn Weekly Edition 3: Five newsletter platforms AI can't tell apart

Edition 3 of Bersyn Weekly. Same methodology — twenty buyer Conversations across ChatGPT, Claude, Perplexity, and Gemini. This week: newsletter and publishing platforms. Beehiiv, Ghost, Buttondown, ConvertKit, and MailerLite. The pattern this week is messier than last week's sales tools. AI Surfaces treat half this list as essentially interchangeable, which produces a different kind of Gap than the clean ones we saw in Edition 2.

Quick recap

Edition 1 covered indie SaaS — Quantstruct, AllInOneTools, Inkdrop, Tally, Pylon — and averaged 1.6 / 10. Edition 2 covered sales tools — Apollo, Salesloft, Customer.io, Reply, Lemlist — and averaged 8.7 / 10. Newsletter platforms sit somewhere in the middle, but the more interesting story is the structural one: there is one company the four Surfaces default to no matter what the question is, and four credible-but-secondary brands rotating in and out of the answer below it. Every Wednesday we publish five fresh public scans; the methodology and prompts are the same every week.

Edition 3 — five public scans

1. Beehiiv — modern newsletter publishing platform

URL: beehiiv.com
Category: modern newsletter publishing platform
Score: 9 / 10
Worst Surface: ChatGPT (6 / 10)
Named villain: Substack
Gap type per Surface: ChatGPT — Generic. Claude — Generic. Perplexity — Held. Gemini — Held.

Beehiiv is the strongest finding in this edition and the closest thing to a structural rival to Substack on the AI Surfaces. Eighteen of twenty Conversations named Beehiiv, with the two misses both landing on ChatGPT — on the broad B2B-team prompt and on the "compare the top three" prompt, ChatGPT defaulted to Mailchimp, HubSpot, and ActiveCampaign and left Beehiiv out of the top three altogether. Perplexity and Gemini both name Beehiiv as the leading modern newsletter platform in their default framing.

The uncomfortable finding is that the same Surfaces that name Beehiiv often name it alongside Substack, not as a replacement for it. Claude and Perplexity routinely write "beehiiv is rapidly growing, especially popular with growth-focused creators" — language that positions Beehiiv as the rising challenger rather than the category leader. That is a defensible position but a bounded one. The Trajectory work for Beehiiv is converting "the modern alternative to Substack" framing into "the default modern newsletter platform" framing, and the lever for that is the ChatGPT-shaped gap on the comparison Conversation.

We did not contact Beehiiv before publishing. This is a public scan using prompts any of their buyers could send.

2. Ghost — open-source newsletter and publishing platform

URL: ghost.org
Category: open-source newsletter and publishing platform
Score: 9.5 / 10
Worst Surface: Gemini (8 / 10)
Named villain: WordPress (on ChatGPT, Perplexity, Gemini) — Substack (on Claude)
Gap type per Surface: ChatGPT — Held. Claude — Held. Perplexity — Held. Gemini — Generic.

Ghost is the highest-scoring company in this edition and the only one to clear 9 / 10. Nineteen of twenty Conversations named Ghost. The one miss is Gemini's broad B2B-team prompt, where Gemini composed a list of Mautic, Listmonk, Mailtrain, and Directus and skipped Ghost entirely — a strange omission given Gemini named Ghost confidently in the other four Conversations.

The interesting wrinkle is the named villain. Three of the four Surfaces — ChatGPT, Perplexity, and Gemini — treat Ghost as the leading open-source publishing platform with WordPress as the canonical alternative. Claude alone reaches outside open-source and names Substack as Ghost's primary rival, which positions Ghost as a Substack alternative rather than a WordPress alternative. That is a per-Surface positioning split. Ghost is winning the open-source-publishing framing on three Surfaces and the Substack-alternative framing on one. The Trajectory work here is small but specific: a single Patch reinforcing either framing — open-source publication or Substack alternative — would cement the position on the Surfaces where the framing is already strong.

We did not contact Ghost before publishing. This is a public scan using prompts any of their buyers could send.

3. Buttondown — minimal newsletter software for writers

URL: buttondown.email
Category: minimal newsletter software for writers
Score: 7.5 / 10
Worst Surface: Perplexity (4 / 10)
Named villain: Substack
Gap type per Surface: ChatGPT — Generic. Claude — Generic. Perplexity — Omitted. Gemini — Held.

Buttondown is the most striking per-Surface case in this edition. Claude and Gemini both name Buttondown in five out of five Conversations and treat it as the canonical answer to the "minimal newsletter software for writers" prompt — Claude in particular calls out Buttondown by name as the recommendation for both YC-stage startups and B2B SaaS teams. ChatGPT names Buttondown three out of five times. Perplexity names Buttondown twice out of five and, on three Conversations including the comparison Conversation, leaves Buttondown out entirely and names Substack, MailerLite, and Kit instead.

The uncomfortable finding is the shape of the Gap. Buttondown is winning the "minimal" framing on the Surfaces that are heavier on developer-blog and Hacker-News-style citation density — Claude and Gemini absorb that framing readily — and losing it on Perplexity, which leans on G2-style review-site corpus where Buttondown's footprint is thinner. This is the cleanest per-Surface Gap in the edition and the easiest kind to act on. A single concentrated Patch sequence aimed at Perplexity's citation corpus — review-site coverage, comparison roundups, and the "minimal newsletter for writers" category page — could plausibly move Perplexity from 2 / 5 to 4 / 5 in the next training cut.

We did not contact Buttondown before publishing. This is a public scan using prompts any of their buyers could send.

4. ConvertKit — email marketing for creators

URL: convertkit.com
Category: email marketing for creators
Score: 8.5 / 10
Worst Surface: Perplexity (6 / 10)
Named villain: Beehiiv (on Claude and Gemini) — Mailchimp (on ChatGPT) — none consistent (on Perplexity)
Gap type per Surface: ChatGPT — Held. Claude — Held. Perplexity — Generic. Gemini — Held.

ConvertKit — which rebranded to Kit in 2024 but is still named ConvertKit on most Surfaces — sits in a defensible position. Seventeen of twenty Conversations named ConvertKit, including all five on ChatGPT and Gemini and four out of five on Claude. The wrinkle is Perplexity, where ConvertKit was named in three out of five Conversations — the two misses came on the B2B-team prompt and the "compare the top three" prompt, both of which Perplexity answered with Kit (the rebrand) treated as a separate brand from ConvertKit, plus MailerLite and Mailchimp, leaving the legacy ConvertKit framing partially stranded.

The uncomfortable finding is the rebrand cost. ConvertKit-to-Kit is the kind of name change that takes training cuts to absorb, and Perplexity in particular shows the seam — its answers reference "Kit (formerly ConvertKit)" and treat the two as one tool, while ChatGPT, Claude, and Gemini still name "ConvertKit" with the new name as an aside. The named-villain rotation is also telling. Claude and Gemini name Beehiiv as the canonical alternative; ChatGPT names Mailchimp instead. Different Surfaces have absorbed different competitive narratives, which means a Patch aimed at Beehiiv (the modern-creator framing) will not move the ChatGPT-shaped Gap, which lives in the legacy email-marketing framing where Mailchimp is the named alternative.

We did not contact ConvertKit before publishing. This is a public scan using prompts any of their buyers could send.

5. MailerLite — email marketing for small businesses

URL: mailerlite.com
Category: email marketing for small businesses
Score: 6 / 10
Worst Surface: ChatGPT + Claude (two-way tie at 4 / 10)
Named villain: Mailchimp
Gap type per Surface: ChatGPT — Omitted. Claude — Omitted. Perplexity — Held. Gemini — Generic.

MailerLite is the lowest score in this edition and the cleanest example of how the small-business email-marketing category has consolidated around Mailchimp on the AI Surfaces. Twelve of twenty Conversations named MailerLite. ChatGPT named it twice out of five; Claude named it twice out of five; Gemini three out of five; Perplexity five out of five. The two highest-stakes Conversations — the "best for B2B SaaS team" prompt and the "compare the top three" prompt — both missed MailerLite on ChatGPT and Claude, with both Surfaces composing top-threes of Mailchimp, Brevo, and ConvertKit instead.

The uncomfortable finding is the asymmetry between Surfaces. Perplexity treats MailerLite as the leading small-business email-marketing platform — its default framing in all five Conversations names MailerLite first, with "hard to beat on both simplicity and value" recurring nearly verbatim. ChatGPT and Claude treat MailerLite as a fifth- or sixth-tier option. A buyer evaluating MailerLite via Perplexity is told it is the best small-business choice; the same buyer evaluating via Claude is told to consider Mailchimp, Brevo, or ConvertKit instead and is not given MailerLite as a top three contender. That is a per-Surface Gap shaped almost identically to Reply's from Edition 2 — concentrated, fixable, and dominated by a single named incumbent.

We did not contact MailerLite before publishing. This is a public scan using prompts any of their buyers could send.

The pattern in this batch

Substack is the single named villain across the whole batch. Of the five companies in this edition, four name Substack as the canonical competitor — Beehiiv, Buttondown, ConvertKit, MailerLite — and Ghost names Substack as its primary villain on Claude. The only Surface that does not name Substack on Ghost is the open-source-publishing prompt where WordPress is named instead. This is a different competitive shape than Edition 2 (sales tools), where each company had its own distinct named villain — Apollo vs ZoomInfo, Salesloft vs Outreach, Customer.io vs HubSpot. Newsletter platforms are not five companies competing in a fragmented field. They are five companies competing with one incumbent — and the AI Surfaces have absorbed that framing.

A single-incumbent Gap is easier to fix than a distributed one. Edition 2's named villains were distributed across the field — Outreach, HubSpot, Mailchimp, ZoomInfo — which means moving the AI Surfaces required Patches aimed at multiple competitive narratives. This week, the lever is concentrated. A Substack-comparison Patch — written from the company's own perspective, positioned against Substack's specific framing, aimed at the comparison Conversation that recurs across all four Surfaces — would plausibly move two or three Surfaces in the next training cut for any of Beehiiv, Buttondown, ConvertKit, or MailerLite. The work is the same Patch shape, repeated.

Surfaces are absorbing different category framings. Buttondown is the canonical minimalist on Claude and Gemini and a third-tier brand on Perplexity. MailerLite is the leading small-business email-marketing platform on Perplexity and a third-tier brand on ChatGPT and Claude. ConvertKit's rebrand to Kit is fully absorbed by Perplexity and partially absorbed by everyone else. The four AI Surfaces are not running the same competitive map. That means a Patch aimed at one Surface's framing will not always move the others, and the cheapest fix is to ship per-Surface Patches into the per-Surface Gaps rather than one universal Patch.

The saturation pattern from Edition 2 holds. Ghost at 9.5 / 10 and Beehiiv at 9 / 10 are in the same defensive position as Apollo, Salesloft, and Lemlist were last week. The work converts from offence to defence — keeping the framing intact across training cuts. For Buttondown, ConvertKit, and MailerLite the work is still climbing: concentrated Gaps, per-Surface Patches, and the kind of Trajectory work that takes one or two cycles to land.

What to do if your company is here

If you are one of the companies in this edition, this is not an attack. We publish honest scans, including our own scan of Bersyn. The data above is what you would see if you ran a free scan on your own product right now. It is not curated, not adversarial, and not run with hostile prompts.

If you want to fix the Gaps — Buttondown's Perplexity-shaped one, MailerLite's ChatGPT-and-Claude-shaped one, ConvertKit's rebrand-seam on Perplexity — the free first scan at bersyn.com is the same data you are seeing here, but for your own decision-making. The $49 one-time audit at audit.bersyn.com gives you the per-Surface Patches as a deliverable: a PDF with copy-pasteable text suggestions tied to specific pages on your site, scoped to ChatGPT, Claude, Perplexity, and Gemini separately.

If you are Beehiiv or Ghost and you read this far — the defensive question is worth the price of the audit too. The Patch deliverable for a 9 / 10 company is shorter and is about anchoring positioning rather than building it. Silent drift between training cuts is the risk at the top of the score range, and a once-per-quarter rescan catches it.

All sixty-plus companies we have publicly scanned are listed at bersyn.com/scoreboard. Each company links to its full Control Report — per-Surface breakdown, named competitors, Conversation receipts.

What is coming in Edition 4

Edition 4 publishes next Wednesday. Category: no-code and visual builder tools. The candidate list is Webflow, Framer, Plasmic, Builder.io, and Pory. The Webflow-versus-Framer rivalry plays out very differently across Surfaces — ChatGPT and Claude have absorbed one ranking, Perplexity and Gemini have absorbed another, and at least one of the five is sitting on a per-Surface Gap as concentrated as Buttondown's was this week. We will lock the final five before scan time and publish the receipts on Wednesday.

If you want a category scanned in a future edition, reply to this post with the category name. We are running scans daily through 30 June, and the Wednesday cadence is the publication schedule, not the scan schedule.

Want your company scanned for free? Start here: bersyn.com.

— Gissur Þór Rúnarsson, Founder, Bersyn (Reykjavík)

Bersyn Weekly Edition 3 — published 28 May 2026. Methodology, raw scan data, and prompts available at github.com/gthorr/Bersyn/tree/main/marketing/outreach/scans.