Explainer
Are AI agents blocked on eBay? Here's what's still allowed
Meta's Muse lost eBay access, but eBay's terms have required permission for AI agents since February. What's still allowed, and the approved way in.

eBay didn’t ban AI agents this month. Its User Agreement has barred buy-for-me agents and LLM-driven bots since February 20, 2026, and what changed in November is that eBay used that rule to cut off Meta’s Muse. If your agent, price tracker or RAG tool touches eBay listings, you need to decide which tasks can stay automated and which go back to a person or through eBay’s own APIs.
My short answer, as of November 16, 2026: checkout is human-only, listing data goes through the Buy APIs, and nothing on eBay is fair game for training. This is research on eBay’s public terms, not legal advice.
Short on time?
Short on time? Here's my take.
- Your agent buys things on eBay
- Hand the user a link to the listing; don't automate checkout
- Both the User Agreement and robots.txt rule out automated ordering. High confidence.
- You need price or listing data
- Apply for the Buy APIs (Browse) through eBay Partner Network
- It's eBay's sanctioned route. High confidence it's the right door; approval isn't guaranteed.
- Your assistant answers 'find me X on eBay'
- Discover listings through a search API, then link out
- Search-engine indexing is the one use robots.txt permits. How eBay views answers built on search results is unstated, so this is my inference. Medium confidence.
- You want to train a model
- Leave eBay data out
- eBay names 'training on our data' as restricted, and robots.txt blocks every training crawler it names. High confidence.
Prices verified . Sponsors cannot buy a verdict or a position.
What happened to Meta’s Muse on eBay? #
Muse users lost reliable access to eBay over the weekend before November 10. Users reported that Muse couldn’t “pull any links or listing details” from eBay, and The Verge reportedly reproduced it. TechBuzz. I couldn’t open The Verge’s article, so this comes through other outlets.
eBay’s statement to The Verge, as quoted by AIChief: “We have restricted agent access in certain cases, including unauthorized scraping and training on our data,” with “ongoing discussions with potential partners”. Meta declined to comment.
A PPC Land roundup names the spokesperson as Sybille St. Arromand. TechBuzz’s line that eBay hadn’t commented is out of date.
On November 12, The Verge called access “somewhat hit or miss”. After its first report, eBay added that it permits “AI search, discovery and certain agent-assisted navigation in accordance with our policies”, per PPC Land. That line has one source and no list of what counts.
The timeline below runs from the terms change to the Muse block.
| Date | What happened | Source |
|---|---|---|
| 2026-01-20 | Updated User Agreement posted, adding “buy-for-me agents” and “LLM-driven bots” | The Register, PPC Land |
| 2026-02-20 | That version takes effect for existing users | User Agreement |
| 2026-06-28 | Current agreement takes effect; clause unchanged | User Agreement, PPC Land |
| August 2026 | robots.txt version labelled v30.2_COM_August_2026 | robots.txt |
| Q3 2026 | Connect deck describes an “Agentic Gateway” for selected external agents | eBay Connect deck |
| Weekend before 2026-11-10 | Muse users lose access to listings | TechBuzz, AIChief |
| 2026-11-10 | The Verge reports; eBay gives a statement | Via AIChief, PPC Land |
| 2026-11-12 | The Verge: access “somewhat hit or miss” | Via PPC Land |
All read and verified 2026-11-16. The Register notes that a general ban on robots and scrapers existed before January; the update added the agent wording, and eBay didn’t comment at the time.
TechBuzz and AIChief both set eBay beside Amazon’s stance on outside agents. I didn’t check Amazon’s side, so I leave it out here.
What do eBay’s terms say about AI agents? #
That any automated access needs eBay’s permission first, whatever it’s for. Section 3, “Using eBay”, bars using “any robot, spider, scraper, data mining tools, data gathering and extraction tools, or other automated means” to access eBay “for any purpose, except with the prior express permission of eBay”. User Agreement.
The parenthesis is the new part: “(including, without limitation buy-for-me agents, LLM-driven bots, or any end-to-end flow” “that attempts to place orders without human review)”. So read-only price tracking needs permission too, not just bots that buy.
The same section bars users from trying to “harvest or otherwise collect or use information about users without their consent” and to “circumvent any technical measures used to provide our Services”. The only exception the clause names is eBay’s “prior express permission”; it doesn’t mention search engines or APIs.
What does eBay’s robots.txt allow? #
More than the terms suggest, for four named assistants. The header says automated access “without the express permission of eBay is strictly prohibited”, says eBay may permit it “solely for the limited purpose of including content in publicly available search engines”, and adds: “Checkouts are strictly for human users.” eBay robots.txt.
The rules then split AI user agents three ways:
| Group (robots.txt v30.2, verified 2026-11-16) | User agents | What it allows |
|---|---|---|
| Training crawlers | GPTBot, ClaudeBot, PerplexityBot, CCBot, meta-externalagent, AmazonBot, Applebot-Extended, Bytespider, ChatGLM-Spider, anthropic-ai | Disallow: /, with small exceptions such as /help and /sellercenter |
| Named assistants and AI search | ChatGPT-User, OAI-SearchBot, Claude-User, Claude-SearchBot | Allow: /sch/ and item pages, but not /itm/addToCart, /itm/watch/, /itm/*action=BESTOFFER or /sch/ajax |
| Everyone else | Any agent not named, including Perplexity-User, Meta-ExternalFetcher, Google-Extended, anything called Muse, and yours | Falls under *, which disallows /sch/ search pages broadly |
The file names 15 AI user agents in total. Muse isn’t one of them.
Here’s the conflict. The agreement requires permission for any automated access; robots.txt path-allows four OpenAI and Anthropic assistants. My read is that robots.txt is an instruction to crawlers while the agreement is the contract users accept, so the agreement is what eBay would rely on.
Those path rules may be the “certain agent-assisted navigation” eBay says it allows, but eBay doesn’t say so; that’s my inference. For the legal background, see is robots.txt legally binding?.

So what is allowed, task by task? #
Very little without permission. The table puts each task against eBay’s two documents. The last column is my reading, not eBay’s.
| Task | User Agreement | robots.txt | My read |
|---|---|---|---|
| Find listings through a search engine or search API, then link out | Search engines not mentioned in the clause | Search-engine indexing is the stated permission | Lowest-risk route (inference) |
| Your own agent reads a listing for a user | Needs “prior express permission” | Not named; * blocks /sch/ |
Not without permission |
| ChatGPT-User or Claude-User reads a listing | Needs permission | Search and item pages allowed | Unclear: the documents disagree |
| Bulk price tracking or scraping | Needs permission “for any purpose” | Unnamed bots fall under *, which blocks /sch/ |
Use the Buy APIs |
| Add to cart, watch, make an offer | Needs permission; order flows “without human review” named | Disallowed even for named assistants | No |
| Check out | Needs permission | “Checkouts are strictly for human users.” | Human only |
| Train a model | Needs permission | Training crawlers get Disallow: / |
No; eBay names “training on our data” |
Sources: User Agreement, robots.txt, eBay’s statement via AIChief. Verified 2026-11-16.
If your assistant needs to answer “find me X on eBay”, I’d use a search API for discovery and send the user to the listing. My SERP API roundup compares those, and search vs fetch vs browser explains why a fetch or a browser is the wrong tool here.
Is there an approved way in? #
Yes, for data: the Buy APIs. The Browse API covers search and item details and isn’t marked Limited. The Deal, Feed (beta), Offer and Order APIs are “Limited release”, “available only to select developers approved by business units”. Buy APIs overview.
Production access isn’t open sign-up. eBay says “The use of eBay’s Buy APIs in production is intended for eBay partners only”; you apply through eBay Partner Network, and “Within 10 business days, the eBay Partner Network will respond”. Contracts depend on your business model, and for member checkout “you need approval” separately. Buy API requirements.
For agents, eBay is building something else. Its Connect 2026 deck, from Nitzan Mekel, VP and Chief AI Officer, says “Agentic Gateway provides a governed entry point for selected external agents and partner experiences”, plus MCP access to “selected eBay data, tools, and capabilities”. eBay Connect deck. The same deck says eBay’s own Shopping Agent launched “to 3% of US native users”.
I found no public application page for the Gateway. PPC Land says eBay hasn’t published which agents have permission or how to get it; the Buy API partner route and the Gateway deck are at least a partial answer.

What I would do #
If I were building on eBay data or agents today:
- Put ebay.com on the “hand back to the human” list for anything past viewing a listing. Show the link; let the person buy.
- Apply to eBay Partner Network if eBay listings are core to your product, and plan for the Browse API rather than page fetches.
- Use search for discovery, not your own crawler, and link out.
- Keep eBay out of training data entirely.
- Don’t route around a block. No proxies, no borrowed user agents. The agreement also bars circumventing “technical measures”.
For the wider legal picture, see is web scraping legal? and is scraping for AI training legal?. The agent browsers pillar covers the tooling side.
What I couldn't verify (as of November 16, 2026)
- The Verge’s reports. Not opened; every Muse detail is second-hand.
- How eBay blocked Muse. No method is stated anywhere.
- API License Agreement and Browse API limits. Both pages returned 403.
- The Agentic Gateway. Whether outsiders can apply, and on what terms.
- eBay’s “agent-assisted navigation” line. One source, no criteria.
FAQ
Did eBay just ban AI agents?
No. The User Agreement wording that names buy-for-me agents and LLM-driven bots was posted on January 20, 2026 and took effect for existing users on February 20, 2026. What happened in November is that eBay blocked Meta's Muse under it.
Can ChatGPT or Claude still read eBay listings?
eBay's robots.txt lets ChatGPT-User, OAI-SearchBot, Claude-User and Claude-SearchBot fetch search and item pages, but not add-to-cart, watch or best-offer paths. The User Agreement still requires eBay's prior express permission for any automated access, so the two documents point in different directions.
Is scraping eBay for prices allowed?
Not without permission. Section 3 of the User Agreement bars scrapers and other automated tools 'for any purpose' unless eBay gives prior express permission. The sanctioned route for listing and price data is the Buy APIs, starting with the Browse API.
How do I get access to eBay's Buy APIs?
eBay says production use of the Buy APIs is intended for eBay partners only. You apply through eBay Partner Network, which says it will respond within 10 business days, and access comes with business-model-specific contracts. Member checkout needs separate approval.
Does eBay have its own shopping agent?
Yes. eBay's Connect 2026 deck says its Shopping Agent launched to 3% of US native users, and describes an Agentic Gateway as a governed entry point for selected external agents.
Which wins, eBay's robots.txt or its User Agreement?
robots.txt is an instruction to crawlers; the User Agreement is the contract users accept. My reading is that the agreement is what eBay would rely on, but that's an inference, not legal advice.
Sources
10 sources, accessed .


