Why LLMs rarely cite Substack or Medium (the 0.07% reality)
Substack earns 0.07% of AI citations and Medium 0.36%. Why newsletter and blog-host platforms underperform, and where to publish instead.
Every GEO consultant has told a founder to "start a Substack" or "repost on Medium" for AI visibility. It is intuitive advice: those platforms rank in Google, they read cleanly, and they feel like publishing. The data says it is close to a dead end. Michael Brito of Zeno Group ran 1.7 million unbranded prompts across finance, travel, automotive, fashion, entertainment, food, and B2B SaaS, and found Substack earned 1,140 citations, roughly 0.07% of the total. Medium did 5.4x better at 6,151 citations, and that still only comes to 0.36%. Neither is where AI engines go to answer a question.
That gap is not a content problem you can outwrite. It is structural, and understanding the structure tells you exactly where your GEO effort should go instead. Signals runs an aged Reddit account marketplace plus an editorial network for AI brand mentions across Reddit, Quora, Product Hunt, and Threads, so we treat every publishing surface the same way: as a retrieval source with a measurable citation profile, not a place to post and hope. This piece is the operator read on why Substack and Medium underperform in AI answers, the one nuance that keeps Medium from being useless, and the surfaces that actually earn citations for the same effort.
Does ChatGPT cite Substack or Medium?
Rarely, and this is the correction that stops founders from pouring months into the wrong channel. Across the largest available datasets, newsletter and blog-host platforms sit far down the citation stack. Brito's 1.7-million-prompt analysis put Substack at 0.07% and Medium at 0.36% of all citations. For comparison, Reddit is the single most-cited domain across every major AI engine, and even the top domain rarely exceeds 5% of the total — so a platform at 0.07% is invisible by any practical measure.
The honest read: publishing on Substack is not an AI-visibility strategy, and publishing on Medium is a weak one. If your buyers ask ChatGPT "best analytics tool for agencies," the cited source is far more likely to be a Reddit thread, a Wikipedia entry, or a "best X" listicle than a Medium post or a newsletter issue. Match the surface to where the engine actually looks before you commit a content calendar to it.
How often do Substack and Medium actually get cited?
Not often enough to build on. The clearest numbers come from Brito's study, and the ratio between the two platforms is as instructive as the absolute figures. Medium pulled 6,151 citations to Substack's 1,140 — a 5.4x edge — yet both round to statistical noise against the domains that dominate AI answers.
The table below sets the two platforms against the surfaces that actually carry AI citations, using the largest 2026 datasets available.
| Surface | Citation profile | Source |
|---|---|---|
| #1 cited domain across every major engine | Search Engine Land | |
| Wikipedia | Leads overall at ~3.4% citation share | Peec AI, 30M sources |
| Medium.com (aggregate) | ~#9 domain overall, but only 0.36% of citations | Peec AI / Brito |
| Substack.com (aggregate) | 0.07% of citations | Brito, Zeno Group |
The interpretation matters. Medium appearing in a top-10 domain list is an artifact of aggregate volume — millions of posts across every topic add up — not evidence that your Medium post will be cited. Per prompt, per query, per brand, the odds are tiny.
Why is Substack structurally invisible to AI engines?
Because the platform is built to fight retrieval at almost every layer. Substack optimizes for email delivery and paid subscription, and each of those design choices removes a signal AI engines rely on. It is not that Substack writers are worse; the plumbing simply does not feed the retrieval layer.
Three structural problems, in order of severity:
Paywalls and signup gates. Premium posts sit behind payment, and even free posts often push an email-capture wall. Content a crawler cannot fully read is content an engine cannot fully cite.
Fragmented domain authority. Every newsletter lives on its own subdomain (
yourname.substack.com). Authority never consolidates into one strong domain the way it does for Medium.com or Wikipedia.org, so no single newsletter accrues the domain-level trust that predicts citation.Email-first distribution. Substack rewards inbox opens, not search indexing. The growth loop routes around the web index that RAG-based engines retrieve from.
Add those up and 0.07% is not surprising. It is the expected output of a platform that treats the open web as secondary.
Why does Medium get cited more but still not enough?
Because Medium solves Substack's domain problem and none of the ownership problem. Every Medium post publishes to one unified, aged, SEO-optimized domain with real author profiles and clean, extractable formatting — exactly the signals retrieval engines and their crawlers reward. That is why Medium clears Substack by 5.4x and lands in top-10 aggregate domain rankings.
The catch is that all of that authority is Medium's, not yours. When ChatGPT cites a Medium URL, the entity it reinforces is the author and the claim, on a domain you rent and cannot control. You cannot add schema, you cannot own the canonical, and you compete against millions of other Medium posts for the same retrieval slot. The result is a channel that can produce the occasional citation but never compounds into brand-level authority. It is the difference between a mention that builds your entity and a mention that builds Medium's.
0.36%
Medium's share — 5.4x Substack, still a rounding error against Reddit and editorial domains.
Source:::
Should you ever publish on Substack or Medium for AI visibility?
Yes, but only as a tier-3 supporting move, never the foundation. There is a real nuance that keeps Medium off the zero line: for a brand whose own site has almost no domain authority, a clean, well-structured Medium post can out-cite the identical article self-hosted, because it borrows Medium's crawl priority and trust. New brands with thin sites can use that on purpose.
But borrowed authority is exactly the ceiling. The durable play is to earn mentions on domains that already carry retrieval weight and to feed the community and reference surfaces engines lean on — the pattern our 50-domains analysis maps in full. Lead with editorial brand mentions, because unlinked mentions correlate 3x more strongly with AI citations than backlinks, and seed the community layer the way how LLMs use Reddit describes. Treat Substack and Medium as repost distribution, not as the place your citations come from.
Frequently asked questions
Does ChatGPT cite Substack newsletters?
Very rarely. In Michael Brito's 1.7-million-prompt study across eight industries, Substack earned 1,140 citations, about 0.07% of the total — a rounding error against Reddit, Wikipedia, and major editorial domains. Substack is built for email delivery and paid subscriptions, so paywalls, signup gates, and its one-subdomain-per-newsletter model all work against the retrieval signals AI engines rely on. If AI visibility is the goal, a Substack is not the channel to build it on. Use surfaces engines actually retrieve from, like community threads and high-authority editorial placements.
Is Medium better than Substack for getting cited by AI?
Yes, by roughly 5.4x, but both are weak. Medium earned 6,151 citations to Substack's 1,140 in the same study, which is 0.36% versus 0.07% of all citations. Medium wins because every post publishes to one unified, aged, SEO-optimized domain with clean formatting and real author profiles — the signals crawlers and retrieval engines reward. Substack fragments that authority across separate subdomains. The important limit is that Medium's authority belongs to Medium.com, not to you, so citations there build the platform's entity more than your brand's.
Why do LLMs rarely cite blog-host platforms like Substack and Medium?
Two reasons: access and ownership. Substack hides content behind paywalls and email walls and splits authority across per-newsletter subdomains, so little of it reaches or accrues in the retrieval layer. Medium reaches the retrieval layer fine but concentrates authority on its own domain, where your individual post competes against millions of others for a citation slot you cannot optimize with schema or a canonical. Meanwhile AI engines concentrate citations on Reddit, Wikipedia, review sites, and established editorial outlets, which offer either community consensus or independent authority that a blog-host post cannot match.
Where should I publish instead for AI citations?
Prioritize the surfaces engines actually retrieve from. Reddit is the single most-cited domain across every major engine, so community presence there is high-leverage. Above that, earn editorial brand mentions on established, high-authority sites, since unlinked mentions correlate about 3x more strongly with AI citations than backlinks do. Reference surfaces like Wikipedia and structured review sites (G2, Capterra) carry disproportionate weight for brand and category queries. Substack and Medium can serve as tier-3 repost distribution, especially if your own domain is new and thin, but they should never be the foundation of the strategy.
Can a new brand with a low-authority site benefit from Medium?
Yes, and this is Medium's one genuine use case. If your own domain has almost no authority, a clean, well-structured Medium post can out-cite the identical article self-hosted, because it borrows Medium's crawl priority and domain trust. That makes Medium a reasonable bridge while you build your own authority and earn mentions elsewhere. Treat it as temporary scaffolding, not a destination: the citations reinforce Medium and your author profile, not a domain you own, so the long-term work is still to earn placements and community presence on surfaces with independent retrieval weight.
:::
