Do Medium or Substack Articles Help AI Assistants Cite You?

Do Medium or Substack Articles Help AI Assistants Cite You?

November 18, 2026·10 min read

Yes, they can: a public post on Medium or Substack is a web page that the search crawlers behind ChatGPT and Claude are allowed to read, so it can be the page an assistant quotes and links when it answers a question in your field. But the platform, not you, decides which crawlers get in and what surrounds your words, so a platform post works best as a second copy or a summary that points back to the original on a site you own.

Can an assistant actually read and cite a Medium or Substack post?

Start with who does the reading. The AI companies run separate crawlers for training and for search, and they treat them separately. OpenAI's overview of its crawlers says GPTBot collects content for training its models, while OAI-SearchBot is for search, and that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. A third agent, ChatGPT-User, may visit a page when a person asks ChatGPT a question. The two robots.txt settings are independent: a site can keep GPTBot out and still let OAI-SearchBot in.

Anthropic describes the same split in its help article on Claude's crawlers. ClaudeBot collects content that may be used for training; Claude-SearchBot indexes content for search results; Claude-User may access a page when someone asks Claude a question. Blocking Claude-SearchBot, the article says, may reduce a site's visibility and accuracy in user search results.

Now look at the platforms. Medium's robots.txt file (medium.com/robots.txt), as published in November 2026, puts Amazonbot, Applebot-Extended, Bytespider, ClaudeBot, FacebookBot, GoogleOther, GPTBot and meta-externalagent in one group with "Disallow: /", leaving only a few of Medium's own pages such as /about and /membership open to them. It does not name OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User or PerplexityBot. Those fall under the general rules, which close areas like editing pages, internal search results and the logged-in /me/ pages, not the stories themselves. In plain terms: Medium's file tells the training crawlers of OpenAI and Anthropic to stay out, and leaves its stories open to their search crawlers.

Substack's platform file (substack.com/robots.txt), in November 2026, names no AI crawler at all. It shuts out one SEO crawler, BLEXBot, and closes areas such as sign-in, subscribe and comment pages for everyone. Each publication also answers with its own robots.txt at its own address, so check yours, not only the platform's.

Two limits apply on both platforms. First, a paywall or a member-only gate can hide text from crawlers as it hides it from logged-out readers, so count only on the part of a post anyone can read. Second, being readable is not the same as being found. On 19 October 2026 we asked assistants for the exact titles of 18 articles, each older than three weeks, published on three client sites. ChatGPT found 1 of 18; Claude found 7 of 18. Google had indexed 29 of 30 articles of one of those sites. Being in Google's index did not mean an assistant could find the page. So check whether assistants find a post, wherever it lives, instead of assuming they do.

When does a post on Medium or Substack help?

A platform post helps most when it answers one specific question under your name and nothing else on the web answers it as well. Assistants that search build their answers from the pages they find for the question asked. A post titled with the exact problem you solve, written from your own work, is a candidate for that question wherever it is hosted.

It also helps when you have no site yet, or a site with very little on it. Substack's own guide to SEO says Substack takes care of making posts appear in search, sets each post's SEO title, SEO description and URL by default from its title and content, and lets you edit all three. It also says you can use a custom domain or the default yourname.substack.com address, and that the default lets you borrow from Substack's existing SEO reputation.

And it helps as a second place where the same facts about you appear. An assistant that meets the same name, role and topics on your site, your LinkedIn profile and your newsletter has a consistent picture to repeat. For what assistants can read on LinkedIn itself, see Does ChatGPT Read LinkedIn Profiles and Posts?

When is your own site the better home?

For anything you want to be known for over years, your own domain wins on three points.

You control it. On your own site you decide which crawlers are allowed, what the canonical address of each page is, and how pages link to each other. On a platform, those choices are made for every writer at once. Medium's robots.txt is an example: whatever an individual writer prefers, the platform's file decides which AI crawlers reach every story.

It names your company and field. On your domain, each article sits next to your About page, your services and your contact details, and every link inside it can point to another page of yours. A platform post sits inside the platform's navigation, next to recommendations for other writers. An assistant reading your own page finds who you are and what you do on the same site; reading a platform post, it finds a byline.

It stays. An address on your domain survives a change of host, a change of newsletter tool or a change in a platform's rules. A platform URL lasts as long as your account and the platform's policies do.

There is a fourth point that matters for measurement: on your own site you get the server log. Our own course marketplace (pickaclass.com) received 1,918 fetches on content pages by ChatGPT-User and Perplexity-User (sent while an assistant answers a person) and OAI-SearchBot (ChatGPT's search crawler), 28 July – 11 August 2026 (its server log); a client blog of a few dozen articles received 56 fetches by the same agents across 24 different articles. Only the ChatGPT-User and Perplexity-User fetches happen while an assistant is answering a person; OAI-SearchBot crawls for ChatGPT's search. A platform does not hand you that log for your posts, so you cannot see which of them assistants open.

For the wider question of how assistants choose which people to name, see How to Get ChatGPT to Recommend You as an Expert.

Should you publish the same text in both places?

You can, but understand what happens to two copies of one text. Google's explanation of canonicalization says that when it finds pages whose main content is the same or very similar, it clusters them and picks the one that is objectively the most complete and useful for search users as the canonical. A Google Search result usually points to the canonical page. You can state a preference with a rel="canonical" link, but the same page says that a canonical preference is a hint, not a rule. A copy on a large platform domain can end up as the version that is shown, while your original is treated as the duplicate.

A safer order for republishing:

  1. Publish on your own site first, and let it stand for a while.
  2. Put the copy on the platform afterwards. If the platform offers a canonical link setting, point it at your original.
  3. After publishing, open the page source of the copy, search for "canonical" and see where it actually points. Substack's SEO guide lists the SEO title, SEO description, post URL and social preview image as the fields you can edit, and does not describe a canonical setting.
  4. Where you cannot point the copy back, publish a shorter version on the platform: the main argument, one example, and a link to the full article on your site.

The fourth option is often the best one anyway. Two different texts, one short and one complete, give assistants two pages that agree about you instead of two copies competing for the same place.

How do you check which of your pages assistants actually cite?

Use more than one method, because each sees only part of the picture.

Ask the assistants directly. Put the questions your clients ask to ChatGPT, Claude, Gemini and Perplexity with web search on, note every URL each answer cites, and ask more than once, since answers change between runs. Then ask for the exact title of your article and see whether the assistant finds your site, the platform copy or neither. The method is laid out step by step in How to Check What AI Assistants Say About You.

Use Bing's AI Performance report. Microsoft introduced AI Performance in Bing Webmaster Tools as a public preview on 10 February 2026. It shows how often your content is cited as a source in Microsoft Copilot, AI-generated summaries in Bing and select partner integrations, which URLs are cited, and a sample of the grounding queries used to retrieve them. It reports on your own site; a story on medium.com belongs to Medium's site, not yours.

Read your server log. Look for ChatGPT-User, Claude-User and Perplexity-User in the user-agent field. A fetch is not a citation, but it shows that an assistant opened that page while answering someone.

Recheck the platform's robots.txt now and then. If a platform starts to block OAI-SearchBot, OpenAI's rule applies to its pages as to any site: they will not be shown in ChatGPT search answers, whatever their quality.

What should a founder or consultant do with this?

Keep the original of everything you want to be known for on a site you own: an About page that says in one sentence who you are and what you do, and a page for each question clients ask you. Use Medium or Substack for reach, as copies that point back or as shorter versions that link to the full piece. Check what assistants cite, not only what they can read.

If you run a company and want to know whether ChatGPT, Claude, Gemini and Perplexity name it, ItGrows measures that and works to make the company visible in their answers. If your aim is to publish consistently on LinkedIn and X in your own voice, its LinkedIn product drafts the posts, and you approve or edit each one before it goes out.

Frequently Asked Questions

Does Medium block ChatGPT from reading articles?

Medium's robots.txt, as published in November 2026, tells GPTBot, which OpenAI uses for training, to stay out. It does not name OAI-SearchBot or ChatGPT-User, the agents OpenAI uses for search and for fetching pages while answering, so Medium stories stay open to them under the general rules.

Is it bad to republish my blog posts on Medium or Substack?

It is not forbidden, but search engines treat the two copies as one piece of content and pick one to show. Google says a canonical link is a hint, not a rule, so publish on your own site first, point the copy back where the platform allows it, or post a shorter version that links to the full article.

Should I use a custom domain for my Substack?

Substack says the default yourname.substack.com address borrows from Substack's existing SEO reputation, and that a custom domain is an option. A custom domain keeps the address under your name if you ever move your newsletter, which matters for pages you want assistants to keep finding.

How can I tell whether an assistant cited my Medium post or my own site?

Look at the link in the answer: it shows the domain that was cited. For your own site, Bing Webmaster Tools' AI Performance report lists the cited URLs from Copilot and Bing's AI summaries, and your server log shows when ChatGPT-User, Claude-User or Perplexity-User opened a page.

Read next

A different question worth asking

People increasingly ask an assistant instead of a search engine, and it answers by quoting somebody. We measured our own logs: AI crawlers read our sites roughly 370 times more often than Google does.

Why assistants name some companies and not others →

AI visibility guides

How ChatGPT, Claude, Gemini and Perplexity find and name a business.

All AI visibility guides →