Thirty days after announcing it, TPT has published no AI content policy at all: no clause, no scope, no effective date, and the only AI text it has published in the window is a training-data promise pointing the other way
The question
"What does Teachers Pay Teachers' forthcoming AI ban actually cover, and does it have an effective date?"
Open follow-up carried forward from [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]], which found the ban existed only in trade press. This run is the delta since then, re-checked against TPT's own published text.
What we already know (from the vault)
- The parent brief recorded the ban as trade press only, and flagged it as such. [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]], observed 2026-10-31: "TPT's Seller Guidelines contain no AI clause... Chalkbeat (2026-10-15) reports TPT plans to ban AI-generated resources, quoting spokeswoman Sarah Lehman that seller material 'should be unique, expertly crafted, and AI-free.' No effective date published."
- The stake is market structure, not compliance. Scribble Works sells direct and is not listed on TPT, so no TPT rule binds RDCO. The parent brief set the frame: "Scribble Works sells direct and is bound by neither, so this is not a compliance problem today. It is a signal, not a settled norm."
- Etsy is the contrast case and it is genuinely published. Etsy's Seller Handbook article stamped 2026-10-22 carries a categorical, quotable ban ("AI-generated functional instructions are not allowed on Etsy") naming patterns, recipes, project plans and instructional guides. That is what a written marketplace AI rule looks like, and TPT has not produced one.
- The product's existing answer to the quality objection is the critic gate. Ruling 14 ([[2026-08-31-product-model-rulings]]) requires generated pages to pass the critic before a child sees them, and the parent brief concluded "the parent-trust framing, not the capability framing, is what survives this norm."
What the web says
- Answer to the headline question: there is still no AI clause in any TPT published policy text, as of 2026-11-07. Read in-browser today: the Seller Guidelines (help article 360042626591) run eleven numbered rules and contain zero occurrences of "AI", "artificial intelligence" or "generative"; the Content Guidelines (help article 360046747572) likewise contain none; the Community Guidelines URL (
teacherspayteachers.com/Help/Community-Guidelines/Guidelines-for-All-TpT-ers) now 301s into help article 360043018571 and contains none. (V) - The Terms of Service has not been touched since before the announcement.
teacherspayteachers.com/Terms-of-Serviceis stamped "Last updated: 12/08/2025", runs 99,764 characters, and contains zero occurrences of "AI", "artificial intelligence", "generative" or "machine learning". A policy announced in October 2026 has not reached the contract. (V) - A full-text sweep of TPT's entire help center returns exactly one AI article out of 419, and it is not about content. Querying TPT's own Zendesk article API across all 419 published en-us articles, one matched on title or body: "Will TPT use my resources to train AI?" (article 55110176560276). (V)
- That one article is new since the parent brief, and it points the opposite way from a ban. Created 2026-10-29, last updated 2026-11-04, full text verbatim: "We won't use your resources to train AI that generates teaching materials or other content that competes with your work without your permission. That applies to TPT and IXL's other products and tools, and it applies to outside AI companies too." So the only AI text TPT has published in this window is a training-data commitment to sellers, not a content rule. (V)
- No effective date exists in any TPT-published text, and the nearest thing to one is an explicit deferral. Chalkbeat (2026-10-15) gives no date, describes the policy only as forthcoming and says TPT "plans to ban" AI-generated resources, and does not say the policy text has been published; the only TPT-attributed quotes are Lehman's statement that material "should be unique, expertly crafted, and AI-free," that the policy will support educators "in keeping teachers' expertise at the heart of the marketplace," and that "we're building tools that will speed up our ability to identify and remove AI-generated resources from TPT." (P on the Chalkbeat reading; Lehman's words are a statement to press, not published policy.)
- The only scope statement traceable to TPT's own mouth is a seller blog's transcription of an Oct 8 town hall, and it is narrower than "AI-free." Lit & More (2026-10-17) reproduces TPT as saying "generative AI should not be used to create images or text in resources or product listings on TPT," and that this "includes resource files, cover images and thumbnails, preview files, or product descriptions." On already-listed work: sellers who have published AI-generated content should "please replace it with your own content." On timing: "You can expect to hear more from us in the next few weeks about how we're approaching this change." On enforcement: "We're not going to carelessly take action against proven stores" and "We are targeting the lowest quality resources that detract from TPT." (U — the blog author's own reproduction; no recording or TPT document is linked, so the wording could not be checked against an original.)
- No disclosure or labeling regime has been offered anywhere, by anyone. Neither Chalkbeat nor the town-hall transcript mentions a disclosure or labeling alternative; both describe removal and replacement only. The single adjacent mechanism is a promised reporting feature, pointed the other way: "In the coming weeks, we'll be adding an option to report low-quality AI resources to TPT" (U, same transcript), which the blog author notes had not appeared as of writing. (V on absence across the TPT pages checked; P on the trade-press absence.)
- Secondary trade press has already over-claimed, and the parent brief's caution was correct. The Tech Edvocate describes the ban as a policy implemented 2026-10-15. Nothing in TPT's published text supports that; the help center, the guidelines and the ToS all show no clause three weeks later. Treat any "effective 10-15" claim as false. (V as to the absence it contradicts.)
- TPT would not strictly need a new clause to act. Two existing published rules already reach unoriginal content: Seller Guidelines, "You should only sell resources that you've produced or designed yourself," and Content Guidelines, "Don't post resources that aren't your original authorship (your work has to be your own)." Both were live before the announcement and are unchanged. (V on the wording; the inference that TPT would enforce through them is mine.) (P)
Convergences and contradictions
- Convergence: the parent brief's characterization held up exactly. It said trade press only, Seller Guidelines carry no AI clause, no effective date published. All three are still true seven days later and now verified against a wider set of TPT surfaces than the parent brief checked (ToS, Content Guidelines, Community Guidelines, all 419 help articles, the blog). The one thing the parent brief could not have known is that TPT's first AI help article would land on 2026-10-29 and be about training data.
- Contradiction inside TPT's own messaging, and it is the whole scope question. Lehman's "AI-free" to Chalkbeat is a provenance test that would exclude any AI in the workflow, including assistive use. The town-hall wording is an artifact test: generative AI "should not be used to create images or text" in the listed surfaces. The enforcement gloss is narrower still and is a quality test, "targeting the lowest quality resources" while promising not to "carelessly take action against proven stores." Three different lines, none published, drifting from provenance toward quality as the audience narrows from press to sellers.
- Contradiction with Etsy as the category precedent. Etsy wrote its rule, dated it, named the covered formats and published a rationale inside eight weeks of controversy. TPT announced first and has written nothing in thirty days. The parent brief read both October moves as one norm forming; on closer look only one of them is a norm, and the other is an announcement.
Synthesis for RDCO
The clean answer to the question asked is: nothing, and no. TPT's forthcoming AI ban covers nothing in published policy terms, because thirty days after the Oct 8 town hall there is no AI clause in the Seller Guidelines, the Content Guidelines, the Community Guidelines, or a Terms of Service still stamped 12/08/2025, and exactly one of 419 help-center articles mentions AI at all. There is no effective date, and the only TPT-sourced statement about timing is a deferral. That is a complete answer to the backlog item, and the useful move now is to stop treating the ban as a known rule in our own positioning, because it is not one.
The more interesting finding is what TPT did publish. Its single AI help article, created 2026-10-29 and updated 2026-11-04, is a promise not to train competing models on sellers' resources. In the exact window where a content ban was expected, the publishing energy went to reassuring sellers about extraction rather than to defining the prohibition. Read alongside the enforcement language from the town hall transcript, "targeting the lowest quality resources" and "not going to carelessly take action against proven stores," the shape that emerges is not a provenance ban at all. It looks like a quality-signal enforcement regime announced in provenance language, which is consistent with IXL's pre-existing practice of using "algorithmic tools to identify and demote stores associated with low-quality, AI-generated content." If that is what ships, the operative test is detectable badness, not whether a model touched the file.
That distinction is the one thing here with real consequence for Scribble Works, and it cuts in our favour. A provenance-gated marketplace norm would be existential for an AI-generation product in this category; a quality-gated one is survivable, and ruling 14's critic requirement is already the mechanism that answers it. The parent brief landed on parent-trust rather than capability as the durable framing; this run adds evidence for why that is right, and it is sharper than "trust sells better." The largest incumbent in the category has spent a month unable to write down what AI-assisted even means, publicly oscillating between "AI-free," "generative AI should not be used to create images or text," and "the lowest quality resources." The definitional problem is genuinely hard, and whoever can point at a verification step rather than a provenance claim does not have to solve it. Our claim is checkable where theirs is not: the critic gate is a fact about what the parent receives, not an assertion about what tools we avoided.
Two cautions on using any of this. First, do not build positioning copy that quotes TPT policy, because there is none to quote, and the one thing worse than ignoring the ban would be characterizing it from trade press. The Tech Edvocate already published that the ban took effect 2026-10-15, which TPT's own unchanged pages refute. Second, the absence verified here is real but perishable: TPT said more would come "in the next few weeks," and that window is now open. The right re-check trigger is mechanical and cheap rather than calendar-driven, since the help-center article count for AI moving off one is a single API call and is a far better signal than another round of trade press.
Why this is in the vault
Closes the fifth open follow-up of [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]] with a verified negative, so the Scribble Works competitive read stops carrying a reported TPT ban as if it were a published rule. It also supplies the evidence for preferring the critic-gate/verification framing over an AI-provenance framing in printables-product positioning, since the incumbent's own three incompatible scope statements show the provenance line is not defensible even by the party drawing it.
Open follow-ups
- Does TPT's rule, when written, test provenance or detectable quality? The town-hall enforcement language ("targeting the lowest quality resources", "not going to carelessly take action against proven stores") points at quality while Lehman's "AI-free" points at provenance. This is the only unresolved part that changes RDCO's positioning, and it resolves the moment TPT publishes.
- Has any marketplace in this category adopted disclosure-and-label instead of a ban? Etsy bans with mandatory disclosure for permitted uses, TPT has offered no labeling option at all, and no third model has been found. Worth knowing before assuming ban-or-nothing is the category norm.
Related
- [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]]
- [[2026-08-31-product-model-rulings]]
- [[2026-08-30-step1-memo]]
- [[2026-09-02-tier-line-plan]]
- [[2026-09-19-mentava-motivation-mechanics-paper-translation-scribble-works]]
Sources
Vault
~/rdco-vault/06-reference/research/2026-10-31-marketplace-free-curated-vs-paid-generated-line.md- parent brief, open follow-up #5~/rdco-vault/01-projects/printables-product/2026-08-31-product-model-rulings.md- ruling 14, critic gate~/rdco-vault/01-projects/printables-product/2026-08-30-step1-memo.md~/rdco-vault/01-projects/printables-product/2026-09-02-tier-line-plan.md~/rdco-vault/06-reference/research/2026-09-19-mentava-motivation-mechanics-paper-translation-scribble-works.md- SW-R18
TPT primary sources (all read in-browser 2026-11-07, approx 01:20 ET)
- https://help.teacherspayteachers.com/hc/en-us/articles/360042626591-What-are-TPT-s-Seller-Guidelines - no AI clause; "You should only sell resources that you've produced or designed yourself"
- https://help.teacherspayteachers.com/hc/en-us/articles/360046747572-TPT-Content-Guidelines - no AI clause; "Don't post resources that aren't your original authorship"
- https://help.teacherspayteachers.com/hc/en-us/articles/360043018571--Guidelines-for-All-TPT-ers - Community Guidelines, no AI clause (redirect target of
teacherspayteachers.com/Help/Community-Guidelines/Guidelines-for-All-TpT-ers) - https://www.teacherspayteachers.com/Terms-of-Service - "Last updated: 12/08/2025", 99,764 chars, zero AI mentions
- https://help.teacherspayteachers.com/hc/en-us/articles/55110176560276-Will-TPT-use-my-resources-to-train-AI - created 2026-10-29, updated 2026-11-04; the only AI article of 419
help.teacherspayteachers.com/api/v2/help_center/en-us/articles.json- full-text sweep, 419 articles, 1 AI match- https://blog.teacherspayteachers.com/?s=AI - TPT Blog, no AI policy post; buyer-facing content only
- https://www.teacherspayteachers.com/Press - HTTP 404, no newsroom found on this path
Secondary (trade press and seller blogs, explicitly not policy)
- https://www.chalkbeat.org/2026/10/15/teachers-pay-teachers-to-ban-ai-generated-teaching-resources/ - 2026-10-15; "plans to ban", forthcoming, no date; Lehman "unique, expertly crafted, and AI-free"
- https://lit-and-more.com/teachers-pay-teachers-the-cost-of-exclusivity/ - 2026-10-17; author's own transcription of the Oct 8 Seller Town Hall, no recording or TPT document linked; source of every scope and enforcement quote marked (U) above
- The Tech Edvocate - claims the ban was implemented 2026-10-15; contradicted by TPT's unchanged published pages, do not rely on
Blocked, flagged, not retried
help.teacherspayteachers.comandwww.teacherspayteachers.comreturn HTTP 403 to curl even with a browser user-agent; all primary reads were done via Playwrightblog.teacherspayteachers.com/wp-json/wp/v2/*returns HTTP 403; the blog's own?s=search was used instead- No TPT seller-account surface was accessed and no account was created, per dispatch. Any AI policy communicated only inside the logged-in Seller Dashboard or by seller email would not be visible to this check, and is the one gap in the negative finding.