First Users for an AI SaaS: What 1,400 Impressions Taught
A founder posted seven times on LinkedIn for about 1,400 impressions and almost no engagement. Then 29 Indie Hackers comments said what to do instead.
6 min read · Last updated: October 25, 2026
The fastest way to find first users for an AI SaaS is to show the product working, not to post about it. One founder got about 1,400 LinkedIn impressions across seven posts and almost no engagement on the two they broke down. Then 29 Indie Hackers comments said to demo the product.
Came here from Instagram, LinkedIn, X or YouTube? Here is the full source list you came for. Everything below is built from these, nothing else.
The original post
The product
The benchmark
Key takeaways
- Seven LinkedIn posts produced about 1,400 impressions, roughly 200 per post.
- The first post reached 217 people from 430 impressions and brought 1 profile visit, 0 new followers, 0 reactions and 0 comments.
- One Indie Hackers post asking for advice got 29 comments and 7 upvotes.
- The advice was to show the product doing a real job, such as comparing conflicting sources.
- Impressions are a vanity number. Follow-up questions and product trials are the numbers to track.
What happened with the LinkedIn posts?
A founder building Prizston, an AI platform that helps people research customer problems and test business opportunities, posted seven times on LinkedIn over several weeks. Together the posts earned about 1,400 impressions. That sounds like traction until you read what each post did.
Post 1 got 430 impressions and reached 217 people. It brought one profile visit, zero new followers, zero reactions, zero comments and zero reposts. Post 2 got 270 impressions, reached 97 people, drew no profile visits and collected one reaction. The founder then asked for advice on Indie Hackers, the founder community, and that one post got 29 comments and 7 upvotes.
| Channel | Reach | What came back |
|---|---|---|
| LinkedIn post 1 | 430 impressions, 217 people | 1 profile visit, 0 followers, 0 reactions |
| LinkedIn post 2 | 270 impressions, 97 people | 0 profile visits, 1 reaction |
| All 7 LinkedIn posts | About 1,400 impressions | Views, little real engagement |
| 1 Indie Hackers post | Not reported | 29 comments, 7 upvotes |
The numbers are small, and that is the useful part. Most first-time founders have numbers like these and never publish them.
Why does this matter to a developer building an AI product?
Because an AI product is easy to build and hard to explain. Anyone can wire up a model in a weekend. What is hard is getting a stranger to believe your version solves a real problem, and a text post about general ideas does not do that.
According to Oiti's benchmark of 6,651 LinkedIn posts, updated October 3, 2026, the median post gets 581 impressions, and anything over about 1,600 puts you in the top quarter. Seven posts at about 200 impressions each sit well below that median. The founder was not unlucky. They were an unknown account talking in generalities.
The bigger point is that impressions measure distribution, not demand. A post can be shown to 430 people and still produce no conversation. If you are a developer, you are probably better at building than at broadcasting, so you need a channel where proof beats polish.
What did the Indie Hackers comments say to do?
The commenters pointed at one fix: stop describing the product and show it. The specific suggestion was a concrete demo, such as Prizston comparing conflicting sources on a real question, instead of another post about AI and research in general.
That matches what the product page itself claims to do. Prizston's site describes a workflow of discovery, validation, content creation and learning from audience response, and its Deep Research step is meant to assess demand, competition, alternatives and risks. A demo of that step on a real question is evidence. A post saying the product exists is not.
The page also says the screenshots appear to be illustrative examples rather than real data. That is the gap the commenters were pointing at. Example screens are not the same as a result a stranger can check.
What changes in practice?
The founder's revised plan, as described in the post, has four parts. Each one swaps a vague activity for something you can measure.
- Share real research examples, including the limitations and failures, not only the clean wins.
- Connect with people who are actively validating ideas, instead of posting to a general feed.
- Measure follow-up questions and product trials, not impressions.
- Build one reproducible example that other people can run and judge themselves.
Here is how the old and new approach compare.
| Before | After | |
|---|---|---|
| Content | General ideas about AI research | Real examples with limits shown |
| Audience | A broad LinkedIn feed | People already validating ideas |
| Success metric | Impressions | Follow-up questions, trials |
| Proof | Descriptions | A reproducible example |
Publishing failures is the risky step, and it is also the one that makes the work believable. A reader can trust a tool more when its author shows where it breaks.
How do you find your first users for an AI SaaS?
Start where people already ask for help, and bring a working example. First users rarely come from reach. They come from a small number of conversations where someone sees your product do something useful on their problem.
If you want more examples of the same pattern, read how one builder got a sale before launch and how another reached 40 users with a first app. Both are small-number stories like this one.
What should you do this week?
- Pick one real question your product can answer and run it end to end.
- Record the result, including anything it got wrong.
- Post it in a community where people are doing the thing your product helps with, and ask what is missing.
- Write down your success metric before you post: replies, follow-up questions or trials, not views.
- Reply to every comment, and keep a list of what people asked for.
If you are worried about building the wrong thing, the post on mistakes that kill most AI startups is a useful checklist to read before you ship more features.
FAQ
Are 1,400 LinkedIn impressions across seven posts good?
Not really. It averages about 200 impressions per post. Oiti's benchmark of 6,651 posts, updated October 3, 2026, puts the median at 581 impressions per post. The more important issue is that the two posts with published stats brought one profile visit and no new followers between them.
Why do LinkedIn posts get views but no users?
Impressions count how often a post was shown, not whether anyone cared. Post 1 in this story was shown 430 times to 217 people and still drew no comments or reposts. General posts about AI rarely give a reader a reason to try your product.
Is Indie Hackers better than LinkedIn for early feedback?
In this one case, yes for feedback. A single advice post got 29 comments and 7 upvotes, which is more conversation than seven LinkedIn posts produced. It is one founder's experience, not a controlled comparison, so treat it as a prompt to test both.
What should I measure instead of impressions?
Follow-up questions and product trials. The founder in this story switched to those two. They show that someone moved from seeing your post to engaging with your product.
Pick one real example this week and publish it with its flaws. Then try the free tools on Art of Code to speed up the boring parts, and read how one builder spent 200 days on an AI chatbot builder for the longer version of building before anyone is watching.
Written by Adithya Guttha, Founder of Art of Code.
Sign in to like and save posts.
Keep reading
Discussion
Comments (0)
Want to join the conversation?
Sign in to commentNo comments yet. Be the first to comment!