What €5 a day of X ads bought: cheap clicks, weak engagement

I put two small projects on a €5-a-day X ads budget: my dog training apps directory and PDFCheck. The clicks were cheap. The engagement made me suspicious.

This started as a “what do you get for $5?” experiment. The actual ad account bills in euros, and both campaigns have a €5 daily budget. These are early results from 20–21 October 2026, with 21 October still in progress, not two completed €5 tests.

The X ads report

The two campaigns generated 194 likes and 2,299 native engagements. X reported roughly 0.70% ad engagement for each. Those actions are a different measurement from whether someone uses the product after arriving.

X reported a link click-through rate of 0.2012% for the dog directory and 0.3196% for PDFCheck. Getting someone to click cost roughly 1.5 to 2.7 euro cents.

Both campaigns optimise for link clicks and pay per impression. That setup is buying traffic; I still need to check whether the people arriving do anything useful.

The ads that ran

The dog ads used colourful app marketing artwork. PDFCheck’s ads used AI-generated report scenes and carried X’s “Made with AI” label. These are screenshots of the published ads, not screenshots proving the product works.

See the other four ad creatives

Creative screenshots were taken after the campaign snapshot. Their per-post views and reactions can include different traffic and times; I do not add those counters to the campaign totals.

PDFCheck got visitors. Very few went further.

PDFCheck’s ads had campaign tags, so I could isolate their traffic in Plausible. The aim-pdf-001 campaign recorded 214 visitors, 227 visits and 251 pageviews across 20–21 October, read around 14:27 Europe/Vilnius.

Average tracked visit duration was 38 seconds. At the recorded spend, the cost was about €0.034 per tracked campaign visitor. That is a decent number of visitors for such a small budget.

The goal report was much less encouraging. Two visitors generated four analysis-page visit events, two saw an upgrade offer and one visited pricing. Those groups can overlap; an analysis-page visit does not prove someone completed an analysis.

No signup-completion, upgrade-click or checkout-start goal appeared in the campaign-filtered report. These are tracking observations, not a reconciliation of customer accounts or payments.

The dog directory has an attribution problem too

The dog site recorded 236 visitors, 249 visits and 328 pageviews during the same calendar window. Its overall bounce rate was 90%, with a 21-second average tracked visit duration. Just one visitor triggered an outbound click to an app.

But the original dog ads had no UTM tags. Plausible labelled 234 visitors Direct / None, one Twitter and one Google. I cannot call all 236 visitors paid X traffic or calculate a trustworthy paid conversion rate from that total.

There is another complication: the dog campaign initially ran worldwide. On 20 October, around 20:00 Europe/Vilnius, I narrowed it to the US and 29 European countries. Before that change, X reported 140,100 impressions and 300 link clicks for €2.14.

PDFCheck started later with the narrower country list. The full campaign totals mix different audiences and running times, so they do not establish which product performs better.

It looks bot-like to me

My conclusion so far: X brings a lot of cheap traffic, but the engagement is abysmal. Much of it looks bot-like to me. Watching the visitor counter rise is satisfying until I look at how few visitors take the next step.

I cannot prove bots from these aggregate reports, let alone say all the traffic is automated. Low-intent people, accidental clicks, targeting and the landing pages could also explain poor results. Bounce rate describes tracked behaviour; it is not a bot detector.

The useful result is narrower: cheap clicks have not yet given me convincing evidence of useful product engagement. X’s link clicks and Plausible’s visitors are different measurements, and I am not treating the gap between them as a bot count.

What I would check before spending more

So far, we’re making negative money. We have paid for traffic, and I have not verified revenue attributable to these campaigns. More visitors have not yet demonstrated a return.

Next I want to test $5 on Facebook, $5 on Google Search and a $5 influencer outreach campaign. That is a planned $15 comparison budget, not three campaigns already launched. The X results above remain in euros; each new test will record its actual billing currency.

Those are hypotheses. I do not have measured results for these channels yet. The X benchmark is €0.0151–€0.0273 per reported link click, with weak website engagement; I will compare the new channels on cost per useful action, not ask them to beat X’s raw traffic count.

Each test needs tagged links, one clearly defined audience and landing page, a spend cap, and the same follow-up window. For the directory, the useful action is an outbound click to an app. For PDFCheck, it is a measured completed task, with signup and checkout activity reported separately. I will record visitors, useful actions and verified attributed revenue at 72 hours and seven days after each test ends.