Where else to launch

Two lists. 13 launch directories read against their own pages by hand, with what each one costs and what it demands in return. Then every directory in the list most founders are working from, checked to see which ones still answer.

Read by hand

13 of these 13 were read against their own pages rather than copied from a list. 12 publish a price on their own site; the remaining 1 does not, and this page says so rather than inventing one. 3 require a backlink or badge on your site as a condition of listing. Nothing here is sorted into free and paid, because that would mean parsing someone else's prose — the price sentence under each name is theirs, verbatim. All of it was read on 7 August 2026, and prices change without anybody telling us.

• AlternativeToNot stated link

  Crowd-sourced, so your listing is edited by strangers.

  No listing fee is published. The site describes itself as "a free service that helps you find better alternatives to the products you love and hate".

• BetaListNo-follow linkPrice not published

  Early-adopter audience, but you cannot see the price before signing up.

  They publish no price on their own pages.

• FazierDo-follow linkBacklink or badge required

  Free tier with a 30-day queue, or pay to jump it.

  Basic is free. Lite $29 (shown struck through from $39) and Premium $49 (from $69). Super shows two figures, $119 and $149, and the page does not make clear which is current.

• FuturepediaNo-follow link

  The most expensive listing on this page, and paid-only when checked.

  Verified Listing $497 one-time. The $247 Basic Listing was marked "(Sold Out)" when checked. Enterprise packages are custom-priced.

• Hacker News — Show HNMixed link

  Free, technical, and the comments are the point.

  Free. The Show HN rules mention no cost.

• Launching NextNo-follow link

  Free to submit, with an optional fee to be looked at tomorrow.

  Free to submit. "$99 upgrade" for consideration for publication within 1 business day.

• MicrolaunchDo-follow link

  A queued launch, or $49 to go whenever you like.

  Pro Launch is $49, discounted to $39 with the code shown on the page. A regular queued launch exists but its price is not published, so treat it as unknown.

• Product HuntNo-follow link

  The default launch day for consumer software.

  Free. The launch guide states "Yes. It's 100% free to use."

• SaaSHubMixed link

  Comparison-led directory; listings are moderated on how well you fit.

  No listing fee is published on the submission page. Featured placement exists but is not priced there.

• Startup FameMixed linkBacklink or badge required

  Free listing in exchange for a badge on your homepage.

  Free on the "Verified" plan. Highlight is $19/month and Spotlight is $149/month.

• Tiny StartupsDo-follow link

  Weekly leaderboard, free, aimed squarely at indie makers.

  Free. The site calls itself "a free startup launch platform"; the call to action reads "Submit a Startup — Free".

• Twelve ToolsDo-follow linkBacklink or badge required

  Fast review, and the paid tier is a one-off rather than a subscription.

  Free tier at $0 forever. Pro is a $36 one-time payment, discounted to $25.20 with the code shown on the pricing page — "one-time payment — no subscription".

• UneedDo-follow link

  Queued launches with a do-follow link, free if you can wait.

  Free to join the queue for the next available slot. $29.99 to skip the line and pick your own date. $15 to relaunch an existing product.

Which of the rest still answer

Most founders submitting an AI tool are working from best-of-ai/ai-directories, or from one of the lists copied from it. Nobody who publishes that list checks whether the entries still work, and there is an obvious reason not to: a large share of them do not, and saying so makes the list look worse.

So this is that check. 150 listed URLs across 148 domains, re-parsed from the source list on every sweep. All of them have been probed at least once. 120 answered on the last check. 9 have no DNS record. 1 returned 404 or 410 on the exact page the list points at. 5 refused a bot that named itself, which tells you about their firewall and nothing else. 3 ask crawlers not to fetch that path in robots.txt, so we did not look. 12 did not serve a robots.txt we could read, which left no permission to act on, so they were not fetched either.

19 of the 148 domains sit inside Tranco’s published top 1,000,000. Another 35 are placed somewhere past it — Tranco’s API answers well beyond the list it publishes, and a position several million places outside that list is not the thing people mean by a Tranco rank. Those are shown below and counted for nothing. 94 were looked up and are not ranked at all. That is not the same as dead. Tranco aggregates five providers’ traffic data; a directory can serve real founders every day and never appear in it. What it does mean is that we have no measurement of whether anyone visits — which is why most rows below carry no verdict.

What each column means

Listed as

The name and link exactly as the source list gives them, re-parsed from that repository on every sweep. We add nothing to the population and remove nothing from it.

Submission path

What one HTTP GET to that exact URL did, from a bot that identifies itself as WorthToTryBot and reads robots.txt first. This is a fact about one URL on one date. It is not a statement about the site as a whole, and a page that has moved will show as gone here while the directory itself is fine.

Tranco

The domain’s position in Tranco’s ranking, with the date of the list it came from. Tranco publishes the top one million domains, aggregated from five providers’ traffic data; its API also answers with positions further down. A position past one million is outside the published list and is shown here without being treated as evidence of anything. A rank describes a domain, so where the list points at a page on one — a markdown file on github.com — the number is the host’s and the row says so rather than claiming it. It is not search traffic, not visitors, and not a quality score — and a directory missing from it entirely is small or new, which is most of them.

Verdict

Given only where the two measurements above agree on something a founder can act on. Most rows have no verdict, because most of the evidence does not support one.

How the check is made

• One HTTP GET per listed URL, from a bot that names itself WorthToTryBot/1.0 in every request. It does not pretend to be a browser, which is why some of these rows say we were refused — that is the cost of being identifiable and it is worth paying.
• robots.txt is fetched for every host before anything else and obeyed, including Crawl-delay. A host that asks us not to fetch a path gets no measurement, and that is recorded as an absence rather than a fault.
• One request per host every 1.5 seconds, and no entry is re-checked more often than every 14 days. The whole list is swept over about a week rather than in one burst.
• A single failed request never reaches this page. It takes 2 failures on separate sweeps before anything here says a path did not answer — one timeout is a bad afternoon on somebody else's server or ours, and publishing it as a verdict would be a claim about a business made on one dropped packet.
• Ranks come from Tranco, free and open, and each one carries the date of the list it came from.
• Nothing here is copied from the directories themselves. The name and the link come from the source list; everything else is a log of our own requests.

What the outcomes mean

Answered

The URL in the list returned a 2xx on the last check.

Page gone

The URL returned 404 or 410. That page is not there; the rest of the site may well be.

Refused our bot

The URL returned 401, 403 or 429 to a bot that named itself. Several of the largest directories in this list do this. It tells you about their firewall and nothing else.

No DNS record

The hostname has no DNS record at all. Nothing is served from it.

No answer

The request timed out, was refused, or came back 5xx.

They asked us not to

Their robots.txt asks crawlers not to fetch this path, so we did not. We have no measurement, and that is not a mark against them.

We could not ask

Their robots.txt could not be read — it timed out, or the server answered 5xx. Without it there is no permission to act on, so nothing was fetched. That the file did not answer is suggestive and is not treated as evidence.

What the verdicts mean

Worth an afternoon

The submission page answered on the last check and the domain sits inside Tranco's published top 1,000,000. The door opens and the domain gets real traffic. Neither fact promises you a link, a visitor or a sale.

Skip it

The hostname does not resolve, or the page you would submit on has returned 404 on at least 2 separate sweeps. There is nowhere to submit.

No verdict

Everything else, and it is most of the table. The largest group is reachable but not in Tranco: we know the door opens, we do not know whether anyone walks through it, and no free measurement we have access to would tell us. Saying nothing is the accurate answer, not a gap.

Ordered by Tranco position, unranked last, alphabetically within. The order is a claim too — sorting by verdict would read as a ranking of these directories against each other, and one number is not enough to rank anything.

12 of 150 clear both bars. That is the number worth acting on, and it is small. A Tranco position is not search traffic. When these domains were checked against a paid keyword index on 7 August 2026, two of them held a rank inside the top 650,000 while ranking for 19 and 8 keywords respectively — present in the list, and not worth an afternoon. That measurement cost money, is not repeated on a schedule, and so is not a column here.

If you run one of these and the row is wrong, it is wrong because a request from our bot got that answer on the date shown. Tell us what we should have seen and we will re-check it — contact.

The whole table, with every definition on this page, is at /directories.json under CC BY 4.0. Take it, check it, correct it.

If you are deciding which of these to spend an evening on, the guide to submitting groups them by what each one demands, and this one asks whether any of it is worth it.

The curated list is also available to agents at /api/v1/directories and in /llms-full.txt.