I built a search engine for 700+ AI tools as a side project — 3 months later
I built a search engine for 700+ AI tools as a side project — 3 months later
by Alex from saas.pet
3 months ago I posted here about saas.pet. You can find that thread below.
Quick update on what happened since:
- Reviews: 75 → 245 (all hands-on, all paid by me)
- Best-of guides: 0 → 12
- Comparison posts: 0 → 5
- Alternatives posts: 0 → 5
- Blog posts: 0 → 13
- Categories: 1 → 12
- Daily users: 1 → ~6 (still tiny, but trending right direction)
- GSC clicks: 0 → 1 (yes, still basically 0 — but impressions growing)
What I learned (some of it hard):
- The hands-on paid-testing moat is real, but it's slow. The advice from OlaWebDesigner and aryan_sinh — "playing the long game" and "the moat is trust" — held up.
- My GSC avg position is 15.3. That's page 2. Page 2 gets nothing. The advice from danielabinav16 ("go all-in on comparison queries") and kimani114 ("stop spreading across 75 guides, pick 5-8 closest to page 1") was the most useful thing I read in 2 months.
- The search UX is OK but not the bottleneck. The bottleneck is ranking. I was about to spend 2 weeks rebuilding the search. Now I'm spending those 2 weeks on 5 more comparison posts.
- The community feedback loop is real. 6 comments on the original post shaped what I built next. Thank you all.
What I'm doing next:
- Adding pricing / free tier / no-login filters to search (aman179102 asked, others implied it)
- Writing 20 more "X vs Y for [specific job]" comparison posts
- Tagging every tool with structured
tasksfield (aryan_sinh's question) - Highlighting "personally tested" badges in search results
Numbers I'm not proud of but committing to: daily users 6, weekly clicks 0-2, Google ranking mediocre. But the content is real and growing, and that's the only thing that compounds.
If anyone here has built a similar search/review project and wants to swap notes, I'm around.
Alex from saas.pet
August 31, 2026 (3 months after the original)
3 months in on a side project and still going is already a good sign most people don't get to. What's kept you at it when a lot of side projects quietly die around week 2-3?
Honest answer: the reviews compound. Every tool I test teaches me something about the next tool, and the page-2-to-page-1 push is slow enough that I don't notice the daily. By week 8 I had 75 reviews and a search that worked for me personally; by month 3 I had 245 reviews and people I don't know are finding it.
The project gets easier as it grows, not harder — that's the part that keeps it alive past week 2-3.
That makes a lot of sense. I like the idea that the work compounds over time — not just the reviews, but also what you learn from each one. Going from 75 reviews at week 8 to 245 by month 3 is a great example of why consistency matters. Thanks for sharing the numbers!
Are you tracking searches that return no useful results? Even with a small audience, those could help you choose what to test next. I’d be curious whether people search for tool names or describe the job they need done.
Not yet, but this is the right next thing to ship. Right now I just see the queries that returned something.
I suspect it's mostly tool-name searches ("hex", "cursor", "claude") but the job-based queries ("write cold emails", "transcribe audio") are where the long-tail opportunity is — that's what the comparison posts and best-of guides are for.
Adding a "no results" log to GA4 events is a 1-day change. Will ship it.
u paid to test 245 tools and most of those founders dont know u wrote abt them. telling them is free links and shares, which is what page 2 needs.
You're right. I've only emailed founders when there's a sponsorship angle or they're a friend. That's a mistake — most of them would happily share a "we got reviewed on saas.pet" link if I just asked.
Adding this to the workflow: after publishing a review, email the founder a 1-line "you're live here" note. Pure no-ask, just FYI. Adds maybe 10 min per review and gets me free social shares + links.
You're right. I've only emailed founders when there's a sponsorship angle or they're a friend. That's a mistake — most of them would happily share a "we got reviewed on saas.pet" link if I just asked.
Adding this to the workflow: after publishing a review, email the founder a 1-line "you're live here" note. Pure no-ask, just FYI. Adds maybe 10 min per review and gets me free social shares + links.
Appreciate the raw transparency on the traffic numbers—6 daily users at month 3 is so real for a side project.
Thanks. The hardest part of writing that update was not the work itself but admitting the numbers out loud. Most indie posts either oversell or don't share at all — both unhelpful.
The truth is the trajectory matters more than the snapshot. From 1 to 6 daily users with no paid acquisition is a real curve, and the next 3 months should show whether the SEO + AI-engine work compounds.
The personally-tested badge is your real moat, but AI tools ship weekly — a review from three months ago can quietly be wrong today. A tested-on date (and which version/model you ran) next to the badge would keep the trust signal honest as the tools underneath it change; that's the kind of detail that makes the badge feel like reporting, not marketing. One more thing: with ~6 daily users, interviewing all of them beats any feature round — at that stage a single conversation usually rewrites the roadmap faster than analytics.
Great points — both. Tested-on date with model/version is exactly the fix I should ship next; it turns the badge from marketing into reporting.
On interviews: 6 daily users is the cap, agreed. I'll reach out once I have a stable weekly cohort. If you ever want to swap notes on the directory vs AI-engine channel, I'm around.
Building an AI tool search engine as a side project is a fast way to capture initial search traffic, but long-term success hinges on staying relevant after the hype cycle fades. Without proprietary indexing algorithms or deep monetization loops, pure directory aggregation risks being commoditized as users turn directly to major AI platforms to discover new solutions.
Fair concern — pure aggregation does risk commoditization. The angle I'm betting on is personal testing data that AI engines can't replicate (see @james_utilityseo's note above on directory citations in AI answers).
Where do you think the defensibility actually lives for a directory like this?
Two of your next actions pull against each other. Pick five to eight closest to page one was, by your own account, the most useful thing you read in two months, and writing twenty more comparison posts starts twenty new pages at zero. One of those is the advice you said worked.
Separately, there is a channel you are not measuring at all. We had our own domain run through six AI engines today and every third-party citation that came back was a directory or aggregator page rather than an article. Directories are what those models reach for on best X questions, which is exactly the query your site exists to answer.
So the cheap test before the twenty posts: ask Perplexity and ChatGPT the list question in your strongest category and see whether saas.pet turns up in the sources. If it does, you have distribution Search Console will never show you, because an AI answer frequently sends no referral at all. If it does not, that is a different fix from more posts.
Sharp observation — and you nailed the conflict (I did say pick five, then turned around and started drafting twenty).
I ran your cheap test just now: asked Perplexity "best AI tools for note-taking" and "best free AI coding assistants" — saas.pet came up in 1/2 sources. So the channel is real, just thin. Going to track citations as a metric alongside GSC clicks now.
Will share the four-week delta in this thread. Thanks for the lens.
The testing moat is real effort, but do users actually choose or engage with tools more because they’re personally tested, or is ranking still doing all the work?
Honest answer: I don't know yet, and that's the test I should be running.
6 daily users, 245 reviews, 1 GSC click/28d — not enough signal. But my Qdrant LinkedIn post got 70 views from "I tested this" framing, not Google. So testing moat probably drives retention + shares, not discovery.
The experiment I'd run: cohort-split new reviews, half with "tested-on date" badge, half without. CTR + return-visit after 60 days.
Want to help me design it?
That’s an interesting signal. I’d be curious to see whether the testing angle actually changes behavior. If you’re open to it, what’s the best email to reach you on?
Email's alex@saas.pet — yes, let's continue there.
If you have thoughts on the tested-on date cohort split (half with badge, half without, measure CTR + return-visit after 60 days), I'd genuinely like to hear them. That's the experiment I'm planning to run.
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.