The Moment AutoValuta Is Built For
You are looking at a used car on AutoScout24, or standing in front of one at a private seller. You have the plate. You want to know: is this car what the seller claims it is? Is the asking price reasonable? Are there comparable cars for less money right now? Is it even insured?
AutoValuta answers all of these questions from a single Telegram message — a photo of the plate, or the plate number typed directly.
The Pipeline
Step 1: Plate Reading
Send a photo of an Italian license plate. A vision-language model reads the plate number and simultaneously assesses the visible condition of the vehicle — useful context for the valuation. The bot accepts both compressed Telegram photos and uncompressed image documents.
If the plate is unreadable — obscured, at an angle, poor lighting — the vision model falls back to visual identification: it attempts to determine the make, model, approximate year and fuel type from the car itself, and continues the pipeline on that basis.
You can also skip the photo entirely and type the plate directly. The bot validates Italian plate format (two letters, three digits, two letters) and routes it straight to the lookup.
Step 2: InfoTarga Registry Lookup
The plate is sent to the InfoTarga API, which queries the Italian vehicle registry. The response includes everything officially recorded about that specific car:
- Brand, model and full configuration description
- Registration year and date
- Registration province
- Engine specs: fuel type, displacement in cc, power in kW
- Vehicle use classification
- Insurance status: whether the car is currently insured, which company, and when the policy expires
The insurance check alone is worth the lookup. An uninsured car cannot legally be driven and creates significant liability for a buyer who does not catch it. AutoValuta surfaces this immediately.
Step 3: Market Data from AutoScout24 and Subito.it
With brand, model, year and fuel type confirmed from the registry, AutoValuta scrapes two Italian used car markets in parallel.
AutoScout24 is Italy’s largest structured car marketplace — professional dealers and private sellers with detailed listings. AutoValuta retrieves comparable cars filtered by make, model, year and fuel type, and calculates minimum, maximum, average and median prices across the listings.
Subito.it is Italy’s general classifieds platform — more informal, faster-moving, often with private sellers who price lower than dealers. The same statistics are pulled from Subito listings. Having both data sources gives a complete picture of what the car is actually trading for across the full Italian market.
Up to eight sample listings from each source are passed to the analyst, with price, variant description, mileage, year and location — the full context a buyer needs to understand where a specific car sits relative to the market.
Step 4: AI Valuation Verdict
All the data — vehicle specs, market statistics, sample listings, visual condition — is sent to a 72B language model acting as a used car market analyst. The model returns:
- GOOD VALUE — priced below market, worth buying at the asking price
- FAIR — priced around market average, reasonable but little room for error
- OVERPRICED — priced above market, negotiate hard or look elsewhere
Along with the verdict, the model provides an estimated fair value range in euros, a brief reasoning that references how mileage, age and trim affect the specific car’s position in the market, and targeted buying tips — common issues with this make and model, what to inspect, how much to negotiate.
What the Response Looks Like
The final Telegram message contains everything in one place:
- Full vehicle identity: brand, model, config, year, fuel, cc, kW
- Plate and registration details
- Insurance status with company name and expiry date
- AutoScout24 price statistics with direct links to comparable listings
- Subito.it price statistics with links
- Verdict with estimated value range
- Reasoning and buying tips specific to the car
A personal dashboard tracks every search with its verdict, allowing buyers to compare multiple cars across a search session or review past lookups.
Why This Works
The value of AutoValuta is not in any single data source — it is in the pipeline that connects them. The Italian vehicle registry tells you what the car officially is. AutoScout24 tells you what dealers are asking. Subito tells you what the private market looks like. The LLM synthesises all three into a judgement call that would otherwise require thirty minutes of manual research.
This is the same architectural pattern as CrateVision for vinyl records: a specific moment of decision, a pipeline that aggregates the right data sources, and an LLM that turns structured data into an actionable verdict. The interface is Telegram because that is where the user is — standing in front of a car with their phone.
Use Cases
Private Buyers
Anyone buying a used car privately in Italy benefits from instant access to registry data and live market pricing. The insurance check protects against unknowingly purchasing an uninsured vehicle. The market comparison prevents overpaying. The buying tips flag known issues with specific models before the test drive.
Car Dealers and Traders
Dealers acquiring stock from private sellers or at auction need fast market valuation. AutoValuta provides a structured price range based on live market data in seconds, with no manual searching required.
Fleet Managers
Companies managing vehicle fleets can use the bot to spot-check market values during renewal cycles or when evaluating disposal of existing vehicles.
Building Something Similar
AutoValuta is one of a growing family of decision-support bots we have built that combine registry or database lookups with live market data and LLM analysis. The same architecture applies wherever there is a moment of decision that requires fast aggregation of structured data — property valuation, equipment pricing, inventory assessment.
If you have a use case that fits this pattern — a specific decision where better data would change the outcome — get in touch. We build custom AI pipelines that connect real-world inputs to actionable outputs.