Guide · Hacker News

Show HN rules and AI search: what to prepare before you post

Updated 2026-11-07 · kagentx

Short answer

Show HN only accepts things people can try themselves: blogs, sign-up pages and landing pages do not qualify, ideally there is no sign-up barrier, the author must be present, and vote asking is not allowed. On 2026-11-07 we read HN's robots.txt and a thread page: there are no AI-crawler rules, and the thread text and comments are in the HTML. How often HN threads are cited in AI answers, we have no verifiable data. What to prepare before posting is the link target: an entry that can be tried directly, a home page that states the definition, an article on how it was made, and an honest comparison page.

What is Show HN actually for?

Hacker News's official Show HN guidelines page (read 2026-11-07; the page carries no date) is quite specific. The core: it must be something others can try themselves.

RuleOfficial meaning (our paraphrase)
AllowedThings people can run on their own computers or hold in their hands; for hardware, a video or detailed article; for books, a sample chapter
Not allowedBlog posts, sign-up pages, newsletters, lists and other reading-only material; these should be regular submissions
Project requirementsNot a quickly generated one-off; you made it yourself and will be around to discuss it; explain how and why you made it
InterfaceIt need not be complicated or slick, and early stage is fine; ideally no barriers such as sign-ups or emails
Not readyDo not post; come back once it is ready. Do not post landing pages or fundraisers
TitleStarts with “Show HN”
Version updates“Foo 1.3.1 is out” is generally not enough; a major overhaul may be fine
Vote askingDo not ask friends to upvote or comment. The general guidelines also say not to solicit upvotes or comments

For indie developers building sites and tools, the easiest traps are the “landing page” and the “sign-up barrier”: if people must sign up before they can try it, or your product is only a page of marketing copy, it does not count as a Show HN under the rules.

Can AI crawlers read Hacker News pages?

We saw no AI-specific restrictions, but one thing needs care. On 2026-11-07 we read HN's robots.txt directly and opened a thread page:

ItemResultWhat it tells you
robots.txtUnder the general group: Crawl-delay of 30 seconds; the disallowed paths are action URLs such as /vote?, /reply?, /flag? and /login; there are no separate rules for GPTBot, ClaudeBot or PerplexityBotThe thread page itself is not blocked by robots.txt
A Show HN thread page (requested as a browser)About 436KB; the title is “post title | Hacker News”; the title link points straight to the external URL with no rel attribute on it; all 323 comments are in the HTML; there is a canonical and no meta descriptionThe text is rendered on the server, not by scripts; comment text is in the page too
Requests using crawler User-Agent strings (from our machine, not a crawler IP)PerplexityBot and Perplexity-User returned 200; GPTBot, OAI-SearchBot, ClaudeBot, Googlebot and bingbot returned 419 (the body was only “Sorry”)Googlebot was refused too, which suggests HN mostly verifies that requests claiming to be crawlers come from official IPs, so this test cannot show that a real GPTBot is refused

The third row is a warning, not a conclusion: testing with curl by using a crawler's name is unreliable for sites that verify the source. Whether real crawlers read your pages must be checked in server logs. We found no verifiable data on how much HN threads make up of ChatGPT or Perplexity answers, so we give no percentage. What we can confirm: the pages are public, the text is rendered in the HTML, and the thread contains your URL.

Which pages should I prepare on my site before a Show HN?

A HN thread page itself holds only the title and link you wrote, plus comments. What visitors, and the AI that reads it, ultimately see is the page your link points to. So prepare these:

  1. 1. An entry point that can be tried without signing up

    The rules say ideally no barriers such as sign-ups or emails. What can be tried is the link target, not a landing page.

  2. 2. A first screen that says what it is, who it is for and the price

    People from HN judge quickly, and the AI reads this same screen. For how to write it, see How to write content AI will quote: before and after.

  3. 3. A page on “how it was made and why”

    The rules ask you to explain how and why. Write it as an article on your own site and link it from your first comment; this is the page that will be cited again and again later.

  4. 4. An honest comparison page

    The most common comment is probably “how is this different from X”. Write it as a page in advance and paste the link when asked. For how to write it, see How to write comparison pages.

  5. 5. Pricing and limits written as text

    The comments will ask about pricing, where data goes and whether it is open source. Put it on the page, not only in comments, which sink.

  6. 6. Confirm crawlers are not blocked by your own site

    However many mentions HN brings, it does not help if pages cannot be read. Check robots.txt with AI crawler check; how to write it is in How to write robots.txt for AI crawlers.

  7. 7. Give links a fixed source parameter

    Only then can you recognise these visitors in GA4; the method is in How do you see ChatGPT traffic in GA4?.

After posting, how do I get AI to remember me?

  1. 1. Stay in the thread and answer questions

    The rules ask you to be around for the discussion. When a reply needs a citation, paste a page on your own site rather than writing it only in a comment.

  2. 2. Turn good comment questions into an FAQ page

    The questions asked most in the comments are your buyers' real wording. Turn them into Q&A on your site, using the question as the heading and answering it in the first paragraph.

  3. 3. Ask AI the same questions after a few weeks

    Ask 3 times, count how often you appear and which URLs are in the sources. How to record is in How do you measure AI visibility? Four metrics and a table you can follow.

How do Product Hunt and Reddit differ from this?

ChannelCore of the rulesWhere to find what this guide lacks
Show HNOnly things people can try themselves, no vote asking, the author must be presentThis guide
Product HuntLaunch day is a burst of exposure; product page and /alternatives/ page structureSee: Does a Product Hunt launch help SEO and AI search?
RedditCommunity posts are often cited by AI; answer honestly under your real identitySee: Why do AI answers love citing Reddit?

Entry points to the other two: Does a Product Hunt launch help SEO and AI search?, Why do AI answers love to cite Reddit? What product teams can and shouldn't do.

FAQ

What are the Show HN rules?

Only post things people can try themselves; no blog posts, sign-up pages, newsletters, lists, landing pages or fundraisers; the project must not be a quickly generated one-off; the title starts with Show HN; do not ask friends to upvote or comment. Source: HN's official Show HN guidelines page (read 2026-11-07).

Can I post a landing page as a Show HN?

No. The guidelines say not to post landing pages or fundraisers; sign-up pages do not count either, since others cannot try the thing first.

Are HN threads cited by ChatGPT or Perplexity?

We found no verifiable data, so this is unverified. What we can confirm: HN's robots.txt has no AI-crawler rules, and the thread text and comments are in the HTML.

Why does requesting HN with a GPTBot User-Agent via curl return 419?

We requested HN from our machine using crawler User-Agent strings: GPTBot, ClaudeBot and even Googlebot returned 419, while PerplexityBot returned 200. Googlebot was refused too, which suggests HN mostly verifies the source IP, so this test says nothing about how real crawlers are treated.

Can friends upvote my post?

No. The Show HN guidelines say not to ask friends to upvote or comment, and the general guidelines say not to solicit upvotes.

Sources

kagentx

Do AI answers
mention you?

An evaluation asks ChatGPT, Perplexity and Claude your buyers' questions (with web search on) and records who they recommend, which pages they cite and whether you are among them. Basic $1, results usually in 2–3 minutes.