Substack Finds Creator Matches—but Keeps the Ranking Logic Opaque

Substack opened Creator Match to all publishers on August 25, 2026, adding a dashboard tool that finds and ranks potential collaborators. The company’s August 25 product announcement identifies audience size, shared interests and predicted mutual promise as matching signals, and confirms that publishers can contact suggested partners directly.

The wider release makes collaborator discovery easier, but it does not establish Creator Match as a dependable growth predictor. Substack has not disclosed how the named signals are weighted, defined what makes a collaboration “promising” or published evidence connecting a higher rank with stronger subscriber results.

Creator Match covers discovery and initial contact

The workflow begins in the publisher dashboard. Creator Match produces a curated ranking of other writers and creators, lets a publisher inspect the proposed matches and provides a route to message one of them.

Its role ends at the introduction. The publishers still need to decide whether to work together and agree on the subject, format, timing and responsibilities. Creator Match does not automatically produce, approve or distribute anything.

That boundary separates it from recommendations and cross-posts. A recommendation places another publication in a reader-facing subscription flow, while a cross-post republishes content through another publication. Creator Match may help publishers begin either kind of partnership—or arrange a guest contribution, podcast appearance or live conversation—but it is not itself a distribution format.

The ranking exposes signals but not their weight

Substack’s public description names similar audience sizes and interests as inputs, then says candidates are ranked by the apparent promise of a collaboration for both parties. It does not explain how interests are inferred, how heavily audience similarity affects the order or whether factors such as publication activity, geography, paid readership and engagement contribute.

An August 11 early-access account documented cards containing a creator’s name, biography, subscriber count, shared subscribers, the audience new to each side and a star rating. The writer reported a one-to-five-star scale and said matches below three stars were not shown during the beta; the broad-release announcement does not establish whether those thresholds remain in place.

Those fields can indicate whether two publications are comparable in scale and how much of their readership overlaps. They cannot reveal why one candidate outranks another. Extensive overlap could signal related interests while leaving a smaller pool of new readers, whereas a large non-overlapping audience could offer theoretical reach without demonstrating interest in the proposed project.

A high position should therefore be read as Substack’s assessment of mutual fit, not as a measured probability of growth. The available materials provide no success rate, confidence score or comparison showing that publishers near the top of the list outperform lower-ranked candidates.

The tool cannot make the editorial decision

A ranked match answers only the first part of a collaboration decision: whether two audiences appear sufficiently compatible to justify a conversation. It does not show whether the publishers share editorial priorities, can agree on a useful project or will reach readers who subsequently subscribe and remain engaged.

The public documentation also leaves rejection controls unresolved. It does not explain whether a publisher can permanently dismiss a candidate, record why a suggestion was unsuitable or influence later rankings through that feedback. Without those details, an irrelevant match could reflect either a temporary suggestion or a persistent weakness in the underlying model.

Nor is there published evidence about discovery outcomes after a match. The platform’s ordering may narrow the search efficiently, but the collaboration itself—its topic, format, placement and execution—can affect results independently of the ranking.

Publishers need outcome data to judge reliability

Testing whether rank predicts growth requires a consistent record of the suggestion and what happened afterward. The useful baseline includes each publication’s subscriber count, the card’s shared- and new-audience figures, its position or rating, and normal referral and subscription performance before the collaboration.

The corresponding outcome record should include the collaboration format and publication date, partner-attributed visits, free and paid subscriptions, unsubscribes, and subsequent engagement or retention over a consistent follow-up period. Recommendations, cross-posts and podcast appearances create different kinds of exposure, so combining them without identifying the format would obscure what the ranking actually predicted.

One productive partnership would show that a particular collaboration worked; it would not validate the ordering across Creator Match. Stronger evidence would require repeated, comparable results demonstrating that higher-ranked candidates tend to generate better outcomes than lower-ranked ones under similar conditions.

The same product update also included audience-specific blocks, but those belong to a separate workflow. Substack’s official Help Center instructions confirm that written, video and podcast posts can show different sections to non-subscribers and to free, paid or founding subscribers. Those blocks control what readers see after reaching a post; they do not demonstrate that Creator Match produced incremental discovery.

Creator Match is now broadly available as a faster way to identify and contact possible collaborators. Whether its ranking can reliably grow a publication remains unknown until Substack explains more of the methodology or publishers accumulate comparable evidence linking rank to subscriber quality and retention.

Also read:

• X’s Creator Connect Moves Campaign Work In-House—But Key Terms Stay Opaque
• YouTube Opens Creator Search Across YPP Markets—but Gemini Remains Opaque
• Facebook Creator Assistant Explains Performance—but Its Advice Is Not a Test