How to market a SaaS or AI app on TikTok and Shorts without ads
What to post for a SaaS or AI tool on TikTok, Shorts and Reels, what 2,389 posts of our own software ads got, and how to measure signups without ad tracking.
By Abondy ResearchPublished 9 min read
In short: Post short screen recordings that show your product doing one job, make many versions of each, put them on TikTok, YouTube Shorts and Instagram Reels, and judge them on signups per day, not views. Expect most posts to get little. Our own software ads, 2,389 posts of ads for Abondy (a web app), had a median of 9 views on the posts we could read, and 2.8% passed 1,000, well below the game clips in our data. A later batch with new videos and captions did better (a median of 35, with 9.6% past 1,000), which is why we'd plan for many rounds rather than one launch video.
This guide is for software sold on the web: SaaS tools, AI apps, developer products. It covers what our data says about software posts, what to post, how to make versions, which app to start on, how many posts a test needs and how to measure signups when there's no ad pixel. For consumer mobile apps, where the goal is an install, see how to promote an app on TikTok and how to track installs from TikTok.
Does TikTok work for SaaS and AI apps?
It can, but our numbers say a software ad starts from a harder place than a game clip. The only software product in our data is our own. From October 22 to November 12, 2026, we ran ads for Abondy through Abondy: 2,389 posts of 1,212 videos, by 628 posters, on four apps.
| Posts | Median views | Mean views | Read at 0 views | Never read | Passed 100 | Passed 1,000 |
|---|---|---|---|---|---|---|
| Our software ads (2,389) | 9 | 97 | 11.6% | 10.4% | 23.0% | 2.8% |
| Other posts in the same weeks, not tagged Roblox (8,587) | 43 | 252 | 8.2% | 14.1% | 40.3% | 10.7% |
| Roblox game clips, Sep 12 to Oct 23 (29,256) | 77 | 518 | 10.0% | 8.1% | 47.1% | 17.7% |
"Never read" means our reader never got a view count back for the post (for example a region-locked TikTok), so we don't know its views. "Read at 0" and "Never read" are shares of all posts in the row; the medians, means and the shares past 100 and 1,000 are of the posts we did read.
Most of the "other posts" (92%) were in campaigns filed under gaming. So in the same weeks, posts of our software ads got lower views than the rest of what went out. We can't say why from this table: the groups differ in their videos, their accounts and their audiences, not only in being software. Read our numbers as one software product's results, not a forecast for yours.
The two batches also differed:
| Batch | Posted | Posts | Never read | Median views | Mean views | Passed 1,000 |
|---|---|---|---|---|---|---|
| First | Oct 22 to Nov 5 | 1,907 | 11.3% | 7 | 72 | 1.1% |
| Second, new videos | Nov 11 to 12 | 482 | 6.8% | 35 | 192 | 9.6% |
The first batch's captions made broad claims ("this Abondy app just does my MARKETING?"). The second batch's were built around one concrete number ("this setup posts 10,000 times a day"). The second batch also ran three weeks later and over only two days. So this shows posts with the new videos got more views, not that the new captions caused it. It's still the pattern we'd act on: change the video, run again, compare.
All our numbers come with the same caveats. The wider sample is mostly gaming clips on small accounts, and a few campaigns dominate it: one brand's campaign accounts for half the posts, and Abondy's own campaigns for about a quarter. View counts are the count when we last read each post. Flagged, rejected and deleted posts are left out, and Instagram counts are a floor.
What should a SaaS or AI app post?
This part is advice, not data.
- One job per video. Show the product doing one thing a buyer wants done: a report built, an email written, a bug found. A tour of every feature is a video nobody finishes.
- Show the result first. Open on the finished thing, then show how it got there. For an AI tool, that's the output on screen in the first two seconds, then the prompt.
- Make the screen readable on a phone. Record at a size you can crop to 9:16, then zoom in on the part that matters. Small dashboard text can't be read on a phone.
- Use a face or a voice when you can. A founder or user saying "I used to spend two hours on this" gives a reason to watch the screen.
- Before and after. The manual way, then the same task in your product. It's the quickest way to show what the product is for.
- Answer comments with videos. A question in the comments ("can it do X?") is the next video.
- Put the name on screen and in the caption. Without a link in the post, the name is how viewers find you.
- Keep it short. YouTube allows Shorts "Up to 3 minutes" (YouTube Help, read November 21, 2026), but a demo of one job rarely needs 30 seconds.
How do you turn one demo into many versions?
A single screen recording can become many tests. Keep the footage and change one thing at a time:
| Change | Example |
|---|---|
| The opening line | "I stopped writing reports by hand" vs "This took 4 seconds" |
| The first frame | The result vs the problem vs a face |
| The caption | A question, a number, a claim |
| The length | A 10-second cut vs a 25-second cut |
| The shape | 9:16 for phones, 1:1 for feeds that show square |
Name every version so you can tell later which one did what. We go into how to run this as a test in ad creative testing without an ad budget, and into how versions get made at scale in AI video variations. What changing the opening words and the frame did to views in our own tests is in UGC ad hooks and versions.
Which app should a software product post on first?
All three short-video apps, with the same videos. In our software ads, each app behaved differently:
| App | Software ad posts | Never read | Read at 0 views | Median views | Mean views | Passed 1,000 |
|---|---|---|---|---|---|---|
| YouTube Shorts | 976 | 2.2% | 17.5% | 3 | 103 | 5.8% |
| TikTok | 620 | 27.6% | 8.5% | 98 | 131 | 0.7% |
| Instagram Reels | 537 | 4.1% | 5.0% | 32 | 78 | 0.6% |
| 256 | 13.3% | 10.2% | 8 | 46 | 0.0% |
TikTok gave the best typical post but rarely carried one far. Shorts was the reverse: most posts got almost nothing, but it had the most posts past 1,000. Reels sat in between. Facebook did least for our software ads. The same split shows up in our game data, which is in TikTok vs YouTube Shorts vs Instagram Reels. Posting one video on all three costs no extra footage.
How many posts does a SaaS test need?
More than feels natural. At our software rate of 2.8% past 1,000 views, 100 posts we could read would give about 3 that pass it. On Shorts alone, at 5.8%, 100 posts would give about 6. One post, or ten, tells you very little, because a typical post gets single-digit views and a few posts get most of them: in our software ads, the top 1% of the posts we read carried 20.6% of all views.
So plan rounds. A round is a few demos, three or four versions of each, posted many times across apps, read after a few days. Keep what worked, drop what didn't, and run the next round.
How do you measure signups without ad tracking?
Organic posts don't come with a pixel. You measure them with a few simple signals read together.
- A tagged link where a link fits. Put a link with UTM tags in your own account's bio and your YouTube descriptions, one per app, so your analytics can tell them apart.
- One question at signup. "Where did you hear about us?" with TikTok, YouTube, Instagram and Other as choices. Many people who saw a post will type your name later instead of clicking.
- Searches for your name. Google Search Console's Performance report has a "queries" view that "groups your data by the search query users typed" (Search Console Help, read November 21, 2026). Watch searches for your product's name during a posting round.
- Signups by day against posts by day. Take your usual daily signups from a quiet stretch, then compare the days you posted. The gap is your lift.
Read posts a couple of days after they go up. In our data, TikTok posts that had 100 or more views when we last read them had a median of 97.7% of their views at our first reading a day or more after posting, so the day a post goes up is close to the day people see it. That makes signups by day a fair match for posts by day.
What does marketing a SaaS without ads cost?
| Route | What you pay | What to watch |
|---|---|---|
| Your own account | Your time | New accounts start small; build it anyway, it's the one you keep |
| Creators making videos for you | Per video | You get the video; it still has to be posted somewhere |
| Clipping | Per 1,000 views | Needs long footage to cut, like webinars or podcasts |
| Paid posts by real people (Abondy) | Per post, about 16¢ to 20¢ on a plan | Views per post are uneven |
We compete with the creator and clipping routes, so read the comparison in best UGC platforms for brands and how much a clipping campaign costs with that in mind. For our own software ads, priced as if every post had been charged at 20¢, the mean of 97 views a post (on the posts we read) works out to $2.06 per 1,000 views. A typical post at the median of 9 views works out to $22.22. Since November 20, 2026, a post with under 100 views 24 hours after its link is handed in isn't charged, so the real cost per view can only be lower.
Where Abondy fits
Upload one screen recording or ad. Our video model, Flagship 1.2, makes hundreds of versions with their own look and caption, and real people post them on their own TikTok, YouTube Shorts, Instagram Reels and Facebook accounts. You see every post and its views, and you pay per post, about 16¢ to 20¢ on a plan, from $20 a month. Developer teams can make and post from code or from an AI assistant through our API and MCP server. More for software teams on Abondy for apps.
How we counted
- Software ads: every counted post from our two campaigns for abondy.com, posted October 22 to November 12, 2026 (New York time). A third, six-post campaign of game content under our name is left out.
- Comparisons: all other counted posts made from October 22 to November 12 that weren't tagged as Roblox, and all other posts of videos tagged as Roblox in our August 29 to November 12 window (they went up from September 12 to October 23).
- Counted posts: live or verified, not confirmed as bought views, not from Abondy's test accounts, pulled November 21, 2026. A few were last read before they were 7 days old.
- Posts we never read: 10.4% of the software-ad posts (27.6% of their TikToks) never returned a view count to our reader. They're left out of every median, mean and share past 100 or 1,000, not counted as zero.
- Timing figure: TikTok posts with 100 or more views and a reading taken at least 24 hours after posting.
Questions people ask
Can you market a B2B SaaS on TikTok?
Yes, as long as each video shows one job your buyer recognizes. Expect most posts to get few views. Our own software ads had a median of 9 views across four apps, on the posts we could read, so plan many posts and read the results together.
Is YouTube Shorts better than TikTok for SaaS?
In our software ads, TikTok had the better typical post (a median of 98 views against 3) and Shorts had more posts past 1,000 (5.8% against 0.7%). Post on both with the same videos.
What videos work for AI apps?
Show the output first and the prompt second, in a screen recording you can read on a phone. One task per video, with the product's name on screen and in the caption.
How do I track signups from organic TikTok?
Use a tagged link in your bio, ask new users where they heard about you, and compare daily signups and searches for your name with the days you posted.
How long before a SaaS sees results on TikTok?
Read each post a few days after it goes up, since most views arrive in the first day or two. Judge a whole round of posts, not one.
Abondy turns one demo into hundreds of posts on four apps. You can try it first.
Read next
- Playbooks · 17 Nov 2026 · 7 min read
How to promote an app on TikTok in 2026
Promote an app on TikTok with short screen-recorded demos, posted in volume across many accounts and judged on installs, with our data on what posts get.
- Creative testing · 20 Nov 2026 · 8 min read
Ad creative testing without an ad budget: organic posts as a test bed
Can organic posts test ad creatives before you spend on ads? What 31,973 posts of versioned videos say about noise, sample size and calling a winner.
- Creative testing · 20 Nov 2026 · 6 min read
AI video variations: turning one ad into hundreds of versions
What makes a video variation read as a new video, which edits don't count, and what 31,973 posts of versioned videos show about why volume matters.
