Linktree's New AI Training Policy: What Creators Need to Know Before July 5
Linktree's new Privacy Notice and Terms of Service take effect on 5 July 2026, and the AI-related language inside them is the reason a chunk of the artist internet has been quietly migrating since early June. Three things change at once: an OpenAI integration that shares profile data with ChatGPT, a brand-new TOS Section 14 covering "Generative AI" inputs and outputs, and an internal-training clause that names Linktree's own algorithms. None of this is hidden. The text is live at linktr.ee/s/privacy and linktr.ee/s/terms today, the deadline is real, and the line creators care about is on the first screen of the privacy notice: "If you do not agree, you should discontinue using Linktree before this date." This post walks through what the AI policy actually says, why the DALL-E framing that went viral in early June was directionally right but specifically wrong, what the EU carve-out actually covers, and what to do in the 13 days between now and the deadline.
If you want the verbatim diff with the prior versions side by side, our Linktree July 5, 2026 Privacy Notice & Terms update is the deeper companion piece. This one is focused on the AI training question specifically. For the full 2026 playbook context (stickers, AI features, and the TOS together), see Linktree's 2026 playbook decoded.
What Changes on July 5, 2026
Two documents get a new effective date: the Privacy Notice (linktr.ee/s/privacy) and the Terms & Conditions (linktr.ee/s/terms). Inside those documents, three meaningful shifts land at once.
- OpenAI integration. Linktree begins sharing basic profile information, a summarized biography, the profile picture, and engagement metrics with OpenAI in response to ChatGPT searches that match your profile.
- TOS Section 14, "Generative AI." A new clause permits Linktree to process and store inputs and outputs (including conversation history) from any AI feature in the product, and to share them with "service providers (including AI model providers)" for improving those features.
- Internal training clause. The privacy notice now explicitly says Linktree uses creator data to "inform, train and improve our algorithms" for recommendation and discovery purposes.
The first screen of the new privacy notice spells out what continuing to use Linktree past 5 July 2026 means in plain language:
"If you do not agree, you should discontinue using Linktree before this date."
That sentence is the operative one for the migration window. Everything below it sits on top of that agreement.
The OpenAI Integration: What's Actually Shared
The OpenAI clause is the one that drove the early-June migration conversation. It sits inside the section of the privacy notice describing who Linktree shares your information with:
"with third party applications and services that integrate with Linktree, such as OpenAI. For example, individuals can search for User Profiles using the Linktree app within ChatGPT, and we may share basic profile information, a summarized biography, profile picture, engagement metrics, and other relevant data in response to those queries."
A few things to read carefully. The integration is named: OpenAI. The product mechanism is named: a Linktree app inside ChatGPT, which lets users search for profiles. The specific fields shared are named: basic profile information, a summarized biography, the profile picture, and engagement metrics. The phrase "other relevant data" leaves the field list non-exhaustive.
Linktree's own public artifacts back up the read that this is a productized integration, not an experiment. The /llms.txt file at linktr.ee/llms.txt describes Linktree's paid plans as including "AI features." There is a public MCP server at mcp.linktr.ee/mcp. Linktree's own profile (linktr.ee/linktr.ee) currently promotes an "AI content & caption generator." The ChatGPT search-result integration is the missing piece that ties those public artifacts to the new privacy notice.
What this is not: training. The clause describes profile data flowing to OpenAI as search-result enrichment, in response to queries inside ChatGPT. It does not describe Linktree handing OpenAI the corpus of all user content for model training. Both of those concerns are reasonable, but only the first is what the document actually permits.
TOS Section 14: The "Generative AI" Clause
The Terms & Conditions add a new Section 14 titled "Generative AI." It applies to Linktree's own AI features, not to OpenAI specifically. The opening framing:
"Linktree may offer artificial intelligence features ('AI Features') that generate outputs ('Output') from information you or others provide and/or information available through Linktree or third parties ('Input')."
The operational clause is the next paragraph:
"To provide and support AI Features, we may process and store Inputs and Output (including conversation history) and share them with service providers (including AI model providers) for providing, maintaining, securing, and improving the AI Features."
"Share them with service providers (including AI model providers)" is the load-bearing phrase. It permits Linktree to pass anything you put into a Linktree AI feature, along with whatever the feature outputs, to an external model provider. The stated purposes include "improving the AI Features," which is broad enough to cover provider-side model improvement.
Today, the AI features in Linktree's own product that this most clearly covers are the caption generator and the AI Insights Chat that has been rolling out since April. Whatever ships next inside the Linktree dashboard with an "AI" label sits under the same clause by default.
See How Linkero Compares
18 content blocks, per-block styling, custom domains, and built-in analytics on every plan.
Create your pageThe Internal Training Clause
Separate from both the OpenAI integration and Section 14, the new privacy notice contains an internal-purposes paragraph that names training directly:
"To develop and improve the Linktree Services, including: To inform, train and improve our algorithms so we can deliver the most relevant recommendations and content to you, including User Profiles that you may be interested in."
This is Linktree training Linktree's own systems on creator data, framed as discovery and recommendation. It is not the same thing as the OpenAI integration, and it is not the same thing as Section 14. It is the third leg of the new AI policy and the one that is hardest to opt out of without leaving the product.
The reading: even before any OpenAI or third-party flow, Linktree's own internal models are now explicitly trained on user data per the new policy text.
The European Union Carve-Out
The new privacy notice contains the same sentence in two different places:
"Certain integrations or data-sharing features may not be available in all regions, including for Users or User Profiles located in the European Union."
The first instance sits inside the OpenAI-sharing paragraph. The second sits inside the broader third-party-integrations paragraph. The reasonable read is that the OpenAI integration, and at least some of the third-party data flows, will not apply by default to EU-located profiles.
This is consistent with how a long list of US SaaS products has handled the regulatory environment since 2025. GDPR, the Schrems III ruling still pending in the European Court of Justice, and the general tightening of US-to-EU data-transfer rules make it cleaner to ship a feature with an EU exclusion than to ship it everywhere and litigate after. Our GDPR link-in-bio Schrems III checklist walks through what this means for EU creators across the bio-link category, not only Linktree.
Two practical implications.
For EU-located profiles: the OpenAI sharing is, by the policy's own terms, not enabled by default. Section 14 still applies in principle, because the AI features it covers are global product features. The internal training clause also still applies.
For everyone else: the carve-out gives EU creators a clearer protection than non-EU creators get. There is no equivalent carve-out for US, UK, or APAC profiles in the text.
How to Opt Out Before July 5
There are three real options between now and 5 July 2026. The cleanest path depends on how much you care about your existing Linktree URL and how much rebuild work you want to do.
Option 1: Migrate to an alternative and delete
This is the actual opt-out. Export your existing Linktree page (Linktree has a built-in data export under account settings), rebuild on a new tool, redirect anyone hitting your old linktr.ee URL via your social bios, then delete the Linktree account before 5 July. Our step-by-step migration playbook covers the timing, redirect strategy, and pitfalls.
Two notes on the timing. First, deleting after 5 July does not retroactively remove data that has already been shared under the new policy. The opt-out is forward-only. Second, the data export is a snapshot, not a continuous sync. Do the export close to the migration, not weeks ahead.
Option 2: Downgrade to free and turn off AI features
Less drastic. The free tier removes some of the paid AI features (AI Insights Chat, AI caption generator) from your account. It does not, on the current TOS text, remove the OpenAI profile-search integration; that one applies based on whether your profile exists and matches a query, not based on plan tier. If your concern is Section 14's AI-feature data flows but you are willing to accept the OpenAI search integration and the internal training clause, downgrading is a partial opt-out.
Option 3: Stay on Linktree and accept the new terms
Reading the actual text and deciding the trade-off is acceptable is a real option. The documents are explicit. Linktree links the prior version next to the new one on both pages, which makes a side-by-side comparison the cleanest possible exercise.
The honest read is that this third option is right for creators whose Linktree page is not a primary part of their professional identity, whose audience is not built around protecting creative IP, and whose engagement metrics are not commercially sensitive.
What This Means By Creator Type
The policy reads differently depending on who you are.
Visual artists, illustrators, and photographers. The profile-picture-to-OpenAI flow is the most directly relevant change. Even though the integration is search-result enrichment rather than image-model training, the principle of profile imagery flowing to a third-party model provider is the question the early-June migration conversation was actually about. For this audience, the asymmetric-cost reasoning typically points toward migrating: the downside of staying is real, the downside of leaving is one rebuild.
Writers, podcasters, and newsletter operators. The summarized-biography and engagement-metrics fields are more relevant than the profile picture. The bio summary is generated on Linktree's side from your provided content, and the engagement metrics include the data you have already opted into via connected social platforms. For this audience, the trade-off is more about brand-association than IP protection.
B2B founders, consultants, and sales pros. Lower data-volume exposure than for the artist audience, but the brand-association question is still real. If your bio link is the page a prospect lands on before a discovery call, "your profile is searchable inside ChatGPT" is either a feature or a non-feature, depending on whether your audience uses ChatGPT for vendor research.
EU-based creators. The OpenAI clause is, by the policy text, not enabled by default. The internal training clause still applies. The Section 14 AI-features clause still applies for any AI feature you use in the dashboard. Reviewing your account settings to confirm what is and is not enabled is the responsible step before the deadline.
Alternatives That Don't Train AI On Your Content
The bio-link category is in active reshuffle. The three questions worth asking of any alternative you are evaluating:
- Does the tool train AI models on user-uploaded content?
- Does the tool share user data with third-party AI providers, either as integration data or as model-training data?
- Where is the tool's data-jurisdiction posture — US, EU, or somewhere else?
A quick read across the major options:
- Linkero. EU-run, no AI training on user content, no third-party model integrations that pass creator data. Bio-link hosting is the only product. Transparent terms. (Full disclosure: this is my product.) For a feature-by-feature comparison, see Linktree vs Linkero.
- Bio.Sites by Squarespace. The product itself is clean, but Squarespace's broader TOS governs it, and Squarespace has its own AI feature roadmap. Not a categorically AI-free option. See how Bio.Sites compares for the longer breakdown.
- Liinks. Smaller-scale EU-friendly operator. Worth checking the current terms individually before deciding.
- Beacons. Actively building AI features into the product. Not the right choice if AI-policy avoidance is the core concern.
- Self-hosted (LinkStack and similar). Technically possible, no third-party model integrations by definition. The trade-off is operational overhead. Realistic for developers, not for most creators.
For pricing on any of these, including Linkero, check each tool's pricing page directly. The category has been volatile.
What We Don't Know Yet
A handful of things the live documents do not answer.
Whether Linktree will publish a public-facing announcement before 5 July. The Linktree blog has been quiet for 39 days as of this writing (last post mid-May 2026), while the legal documents have been updated with the new effective date. The "we changed our terms" notice on linktr.ee's blog index is the only blog-level acknowledgment so far.
Whether the OpenAI integration is opt-in by default or opt-out. The privacy notice describes the data-sharing as happening "in response to those queries" without specifying account-level controls. The reasonable assumption is that the integration is opt-out at best for non-EU profiles. Watch for an updated settings page closer to the deadline.
Whether the recent Trustpilot user complaint about AI "reshaping" user-provided text reflects something already in the product or a misread of the AI caption generator. The new policy clearly permits data flows to AI providers for feature improvement; whether that is currently being used to rewrite user content silently is not something the TOS text alone confirms or denies.
FAQ
Does Linktree train AI on my profile picture?
Not as described in the live privacy notice. The OpenAI clause covers a ChatGPT search integration that shares "profile picture" among other fields in response to user queries, not image-model training. The internal training clause covers Linktree's own algorithms for recommendation, not image generation. Section 14 covers Linktree's own AI features, not external image-model training.
Can I opt out without deleting my account?
The TOS text does not describe a granular account-level opt-out. Downgrading to free removes the paid AI features from your own dashboard, but does not, on the current text, exempt your profile from the OpenAI search integration or the internal training clause. The clean opt-out is account deletion before 5 July.
Does this apply to the Linktree free plan?
Yes. The privacy notice and the new Terms apply to "all new users and to existing users continuing to use Linktree" past the effective date. Plan tier does not change which policy applies; it changes only which AI features you can access inside the product.
What's the difference between this and the DALL-E claim from June?
The early-June Bluesky and Tumblr framing said Linktree was about to feed all uploaded imagery into DALL-E for image-model training. The actual integration is a ChatGPT profile-search enrichment that includes the profile picture (singular) among the shared fields. The 5 July 2026 date is correct, the OpenAI integration is real, profile imagery is included, but the specific "DALL-E training" framing and the "all uploaded imagery" scope are both broader than the policy text. The underlying alarm was directionally right; the specific factual claim was partly wrong.
Are EU creators affected?
Less than non-EU creators, by the policy's own terms. Two clauses explicitly carve out EU profiles from "certain integrations or data-sharing features," including the OpenAI clause. The internal training clause and Section 14 still apply. Reviewing your settings before the deadline is the responsible default.
How do I check what Linktree is sharing about my profile?
Linktree has a built-in data export under account settings. The export gives you a snapshot of what is on your account; it does not, today, describe in-line which fields are being shared with which third parties. For the live policy text, the source documents are linktr.ee/s/privacy and linktr.ee/s/terms.
The Bottom Line
The change is real, it is dated, and the AI clauses inside it are narrower than the viral framing of early June. They are also broad enough that creators whose bio pages carry their professional identity have a legitimate reason to migrate before the deadline. Thirteen days is enough time for a clean export, a rebuild on an alternative, a redirect through your social bios, and a deletion that lands before 5 July. The migration playbook is here; the verbatim diff of what the documents actually say is here.
If you want a bio-link tool that explicitly does not train AI on your content, does not share creator data with third-party model providers, and operates from EU jurisdiction, that is the posture Linkero is built on. The comparison and the sign-up are below.
See How Linkero Compares
18 content blocks, per-block styling, custom domains, and built-in analytics on every plan.
Create your page

