In April I submitted itbroke to 18 startup directories over two days. Ninety days later, my analytics showed three referred sessions for the whole site: one from Bing, one from Baidu, one from Facebook. From the 18 directories, combined: zero.
I published that as the finding. It was the wrong finding, because one product is not a sample.
Running the same measurement across four of my products four months later, the spread is the story:
Thirty-four sessions is not a rounding error for a site that size. It is 24% of everything askbowtie got referred in ninety days — and futuretools.link alone sent 24 of them, more than Bing sent, about a third of what Google sent.
So do directories drive traffic? For one of my four products, yes, and it is one of its best channels. For the other three, not one session. The variable is category, not effort. askbowtie is an AI tool, and the AI-tool directories are alive: people browse them the way they once browsed Product Hunt. uxxray and serpdelta are SEO tooling, submitted to general startup lists that nobody reads.
That is the useful version of this post, and it took being wrong in public to find it.
What I actually did
I didn’t guess at this. I keep a database of 147 launch directories, each one checked for domain rating, whether the submit form is alive, and what it demands. On April 10 and 11 I worked itbroke through the list. On April 12 an agent ran the same three open forms for five more of my products in one sitting (Claude Code drove the browser; I reviewed each submission). Every submission got a status row.
The statuses are the honest part:
Submitted and verified: StartupFA.me (DR 83) and Turbo0 (DR 78) both confirmed the listing after checking the backlink on itbroke.dev
Stuck at needs_backlink: 4 of 18, including Fazier (DR 81) and Dang AI (DR 80), which won’t publish until you link back to them
Submitted into the void: Future Tools took the form and showed no confirmation. Four months on it is the single best-performing directory in the list, at 24 sessions to askbowtie in ninety days. No confirmation email is not the same as no listing
That middle column is the thing nobody tells you. The best free directories gate publication on a reciprocal link. You submit, feel productive, and three of your highest-DR listings quietly never go live because you didn’t add a badge to your footer.
The fix took one code change: a single “Featured on” row in the site footer. One edit unblocks Fazier, Dang AI, Toolfio, Twelve Tools, and TinyLaunch at once. Highest ROI move in the whole exercise.
Why the 147 became 31
The padding in those “100+ directories” lists is real. Of my 147 candidates, only 31 survived vetting:
What got cut
Do
- Paid-only listings dressed as free: Futurepedia starts at $247 per listing, aiwith.me at $19.90
- Sites that aren’t directories at all: TechCrunch appears on half these lists and has no submission form
- Dead or hostile forms: 404s on the submit page, or Aura++, which requires your backlink now and won’t publish you until 2027
- Scanner-blocked but human-fine: G2 and StackShare block bots, so automated services silently skip the two most valuable B2B listings
That last point is the case against paying someone. The bulk services automate what shouldn’t be automated and skip what a human can do in two minutes.
I kept adding candidates and re-checking them after that first pass. The published list now sits at 101 free directories, each one confirmed to have a live submit form and labelled with what it demands before it will publish you.
Which is worth saying plainly, because the original version of this post did not: eighteen directories took me two days. A hundred and one is not an afternoon. It is days of repetitive form-filling, and anyone telling you otherwise has not done it.
What you’re actually buying with that time
Not customers. Links from DR 60 to 90 domains pointing at your product with a consistent name, URL, and description. That consistency is what search engines and AI crawlers use to figure out that your five scattered projects belong to one entity, which I learned the hard way when crawlers couldn’t connect my own five sites to one builder.
Set the expectation accordingly. If you’re pre-launch and hoping directories replace marketing, they won’t; my honest numbers on what actually made money are in I built 4 SaaS products with AI.
And one thing I believe but cannot show you: that the backlinks lift domain authority enough to help unrelated pages rank. It is the reason most people do this, the mechanism is plausible, and I have no before-and-after DR data to put behind it. Treat it as untested here.
Do it yourself, in this order
Work the Launch Directories list in this order:
Several sittings, honestly. A footer badge row. Real submissions with a consistent name, URL, and description behind every one. That is the entity trail from You Are Invisible to Machines, and it is the whole playbook.
Erratum, 2026-08-29: the headline claim was wrong
This piece originally concluded that directory submissions are “worthless as a traffic play”. That was measured on one product and published as a general finding. It is false for at least one other product on the same analytics account, and the original wording is retracted.
What the original said, and what is true now:
| Original claim | Status |
|---|---|
| itbroke: 18 directories, 90 days, zero referral sessions | Stands. Re-checked, unchanged |
| “Worthless as a traffic play” | Retracted. askbowtie took 34 directory sessions in 90 days, 24% of all its referred traffic |
| “Future Tools… I still don’t know if it landed” | Answered. It landed. 24 sessions, the best directory in the list |
| “Not worth $50 to $200” / “never worth outsourcing” | Re-labelled. Opinions, never measured. Marked as such above |
| “One afternoon” | Corrected. The post’s own account says 18 directories took two days. The list is now 101 |
Method, so you can repeat it: first-party analytics, get_referrers over a 90-day window ending 2026-08-29, counting sessions whose first external referrer is a directory domain. Referrer data undercounts by design: a link opened without a referrer header lands in “direct” and cannot appear. Every number here is a floor, including the zeroes.
What I got wrong is worth naming precisely, because it is a cheap mistake to repeat: I had a denominator of one, and a denominator of one is an anecdote wearing a number’s clothes. The fix was not better analysis of the same data. It was more data.
Related: