"Will I Get Ripped Off?": a Turkish developer's side-project AI car-inspection app hit 30k installs on one catchphrase and three TikToks

Kazık Yer Miyim? (Turkish for "Will I get ripped off?") is an AI advisor for Turkish used-car buyers. Feed it a car-listing screenshot, and it tells you in Turkish whether the price is too high, whether there are signs of paint or part replacement, what faults the model is known for, and gives a "ripped-off risk score." The developer is a full-time employee building this as a side project. The product itself is a wrapper around a large language model, and similar tools already exist. It took off on a good name and a few short videos bought with ad money. Below: how it grew, why it has slowed, and whether this playbook travels to other countries.

One video with 2.67M views, and one bad review

In Aug 2026 3, the app's official TikTok account posted a 73-second video captioned "Buying a car and want to make the smartest decision? Ask Kazık Yer Miyim first." By Nov 6, that video had 2.67M views. But the account had only 3,512 followers, and its usual videos drew a few hundred to a few thousand views. The video is labeled "promoted content" on TikTok. The July 23 video with 1.63M views is labeled "paid partnership." Both were pushed with money.

Two weeks later, in Aug 18, the Android version launched. On Aug 24, a four-star Google Play review read: "Came from an ad. Good, even great, just the monthly fee is a bit steep... the AI talks like a real consultant."

Three days later, on Aug 27, another user gave one star: "I uploaded screenshots of two cars. Both were manual with no sunroof. It told me this model's DSG (dual-clutch gearbox) has problems and the sunroof might too."

Put these three things together and you have most of the app: money spent on TikTok, a name everyone understands, users drawn in by the pitch; and an AI that can sound like a seasoned mechanic while confidently making things up.

What it is

Turkish buyers of used cars mostly browse listings on two platforms, sahibinden and arabam. Cars are expensive, and buyers fear three things most: paint that's been redone, parts that have been replaced (boya / değişen), and overpaying. The old fix was to ask a "car-savvy friend," or pay for a third-party inspection (ekspertiz).

The app aims to replace the "ask a friend" step, not the inspection center. Using it is simple and requires no forms: a screenshot of a listing, a photo of a car on the road, or a screenshot from a WhatsApp group can all be uploaded directly, up to four images at once for comparison. The output is a card: how far the price deviates from comparable cars, any mismatches in the description, the model's common faults, and a risk score. Later additions included haggling simulation, "My Garage" (maintenance, inspection, and insurance reminders), and "this month's popular models."

The developer is Güney Ural, who works a regular job and built this as a side project. The iOS version launched in May 24, the Android version in Aug 18, and it's sold only in Turkey. It shipped 17 versions in 5 months.

Why this one took off

The name is the ad copy. "Kazık yer miyim?" is an everyday question Turks ask when shopping. It needs no explanation, you get it on sight, and it sticks after one hearing. The ad copy saves half the work.

It sides with the buyer. The website says it is "independent and not affiliated with any listing platform." sahibinden has its own AI for searching cars and identifying them from photos, but platforms depend on sellers and won't tell buyers "this car is overpriced." That position is one the platform structure can't offer, yet someone has to occupy it.

It spent money on TikTok. The timeline is clear: after the iOS launch in May it stayed quiet. Promoted videos on Jul 20, Jul 23, and Aug 3 drew 360k, 1.63M, and 2.67M views, with a like rate of only about 0.3%, which looks like bought views. The Android version launched on Aug 18 to ride that wave. The iOS privacy labels list "third-party advertising" and cross-app tracking, showing it uses ad attribution.

Someone did the same thing a year earlier, and it didn't take off. EksperGPT launched on Google Play in August 2025 year, doing almost exactly the same thing. It has over 10k installs on Play so far, and the iOS version has 30 ratings averaging 2.8 stars. The difference isn't the model. It's whether someone paid to get Turks to see it, and whether the name was easy to remember.

How far it grew, and why it slowed

Turkey Google Play "Auto & Vehicles" free chart: 10-29 #7,11-06 #11. The iOS version has 11-06 ratings averaging 4.44 stars 152, but went from 10-28 to 11-06 with only 15 more. The developer's personal website says "33,000+ users" and claims the Android version "repeatedly charted #1 in the auto free category"; in public chart data, its best since 10-29 is #7.

The curve's shape tells the story: it was a single burst, not a slow climb. The first two weeks caught the afterglow of the promoted videos, reaching nearly 10k installs. From September on, TikTok shifted to a low-cost video a day tagged #sahibinden #ikincielaraba, drawing a few hundred to a few thousand views, and daily installs dropped to two or three hundred. Without paid promotion, volume returned to a plateau.

Instagram has 8,797 followers and only 27 posts, lots of followers but few posts, likely also driven by ads (inference). No media coverage was found, there's no discussion on X, and searching its name on TikTok turns up only the official account's own videos.

Where the money comes from, and how smoothly it's collected

It asks for money right away. Early on, a paywall appeared almost immediately after sign-up. A one-star review from Sep 11 read: "Scanned 5 listings for 199 lira, 3 listings for 139 lira, what is this, bro? At least give a few free trials." The website now lists a monthly subscription at ₺239.99 and an annual subscription at ₺1,999.99 (labeled "save 31%"), plus credit packs of 1, 5, and 15 uses. Google Play in-app purchases run from $0.59 to $119.99. The web version collects through iyzico, a Turkish local payment processor, bypassing store commissions.

Charging this way makes sense because buying a car is a decision worth hundreds of thousands to millions of lira, so paying a couple hundred lira to "just check" feels cheap.

But from the outside, conversion looks rough:

• Most one-star reviews on Play from late August to September complain about the paywall: "You have to pay just to see if it works," "It charges after one message," "Paid before even trying it."
• Some users say directly, "ChatGPT or Gemini will do."
• Someone paid a year's fee of 2000 lira and couldn't log in (10-03).
• The developer cut prices on Sep 22, raised chat credits to 4 times the original on Oct 25, and added in-app ad slots in the same version; on Oct 30 added "pay to pin your own listing."

Cutting prices, adding credits, placing ads, and selling pins, all within a single month, suggests subscriptions alone weren't enough (inference). It has never charted on the top-grossing list and has never disclosed revenue, so actual income is unknown.

Who it's competing with

Verdict: plenty of people are building this, and building it isn't the hard part. What decides who gets the traffic is who spends money on Turkish TikTok first and who has the more memorable name.

If it were you

B: Worth a try, but build it in another country, and you need to be good at short-video ads. Turkey already has a player holding that spot, and everyone has access to the same models, so rebuilding it in Turkey is pointless. What's worth taking is the formula: a local catchphrase as the name, a high-stakes one-time decision as the core, screenshot input, and pay-per-use pricing.

Who it's for: 1–2 people, one of whom has run TikTok or Meta ads. Native speakers or long-term residents of a used-car market with asymmetric information and one dominant listing platform. Candidates include Poland (Otomoto), Romania, the Gulf states, Mexico, Brazil, Indonesia, and Vietnam (inferred, not checked market by market for existing competitors). Ideally you know cars yourself, or have a friend who does to help verify the known-issues database.

Who it's not for: people who only write code and expect organic app store traffic (the dozen or so similar Turkish apps have stalled at single-digit ratings for exactly this reason); anyone who wants to go head-to-head with the incumbent in Turkey; anyone unwilling to take responsibility when the AI is wrong. Inspection advice is high-risk. One bad answer means a one-star review.

How to build it: start with a web version and an Android app, not iOS. Name it after the worry local buyers voice most often. Don't rely on the model alone for price judgments. Scrape public listings from the local listing platform and compute a price range by model, year, and mileage. Let the AI write the commentary only. Turn known model defects into a lookup table the AI cites rather than invents. The first two analyses are free. After that, sell credit packs of 3, 5, and 15 uses, not a subscription on day one. Subscriptions are what Turkish one-star reviews complain about most.

How to get users: follow its rhythm. Test a dozen or so vertical videos on a small budget, with the content "paste this listing in and see what it says." Pick the one or two that perform, put real money behind them, and time the push to get the Android version live while the buzz is up. After that, post one video a day, tagged with the local listing platform's name and car models, to keep volume steady.

Deadline: launch in one country and finish the first round of paid-acquisition tests before March of 2027. If you wait longer, general assistants like ChatGPT and Gemini will keep getting better at "upload a photo, evaluate the car," and the local listing platform may build its own buyer-side evaluation.

Signs you've misjudged it: (1) you've tested more than 20 creatives and none has brought the cost per paying user below first-order revenue. The business doesn't pencil out in your market. (2) the local listing platform launches buyer-side valuation or risk warnings. (3) the Turkish incumbent stops updating or gets pulled in early 2027. Even the one that worked didn't last, which means willingness to pay can't sustain the ad spend. Conversely, if it starts spending again in early 2027 and installs re-accelerate, that suggests the math works in Turkey, and you should move faster.