About

Most "is this engine reliable?" answers are one person's bad luck, or one dealer's sales pitch.

enginecreep exists to replace that with something closer to a pattern: what shows up again and again, across thousands of owners, official complaint databases, and repair forums, for the specific engine you're about to buy a car around.

Who's behind this

enginecreep is a side project built by a small group of computer science students who got tired of finding out about engine problems after buying a car. The idea is simple: take the information that already exists across forums, owner communities, and official databases, and make it actually usable before you sign anything.

Questions, missing engines, or data that looks wrong? All of us have access to contact@enginecreep.com, and anything reported, whether it's a wrong figure, a missing engine, or a correction, gets looked at and fixed as soon as we can.

Where the data comes from

Owner discussions pulled from relevant subreddits: what actually breaks, when, and how people fixed it.

Specialist forums

Marque- and model-specific forums (e.g. PistonHeads and others), searched for the engine code and its known failure terms.

NHTSA

Official US complaint and recall data: the same database regulators use, not a paraphrase of it.

HonestJohn Real MPG

Real-world fuel consumption reported by owners, compared against the manufacturer's official figures.

Wikipedia

Baseline specs: displacement, power, production years, chassis codes.

How raw data becomes one engine page, and stays accurate

Two-pass collection

The first pass gathers everything findable for an engine. It is read once to pull out the specific terms people actually use for that engine's failures: a part name, a fault code, a nickname. A second pass searches specifically on those terms, so the data set goes deeper than a generic keyword search would.

Structuring, not summarizing

The raw data (Reddit threads, forum snippets, NHTSA complaints, recall text) is processed with the help of an AI model, then checked against the source material with one rule above the rest: nothing gets added that is not present in what was actually collected. The job of both the AI and the review step is to organize what is already there into a consistent, comparable shape, not to add opinion or fill in gaps.

Fixed format, every engine

Severity, repair cost ranges, km ranges, and the overall verdict all follow the same strict format across every engine, so a BMW N47 page and a Ford EcoBoost page are directly comparable, not written in whatever style happened to come out that day.

Sanity-checking the numbers

Reported fuel consumption is checked against what is physically plausible for that engine type. If a figure falls outside what is plausible, or if reported real consumption comes out lower than the official rating, it is marked as unavailable rather than shown as fact.

Read by a person, on a schedule

Every blog post and every engine page is read in full by one of us, checked against the data actually collected, to make sure it still matches reality. How often depends on the engine's age: engines that have been on the market 15+ years, where most of what will go wrong is already known, are re-reviewed every 6 to 12 months. Engines roughly 7 to 15 years old are reviewed every 3 months. Anything under 7 years old, where new failure patterns are still emerging, is reviewed monthly.

What this isn't

This isn't a guarantee, and it isn't a mechanic's inspection of the specific car you're looking at. It's a summary of what's publicly documented about an engine as a model. Your exact example could have been maintained perfectly or neglected for years, and no amount of forum data can tell you which.

Owners who post online skew toward people having problems: nobody starts a thread to say their engine is fine. enginecreep tries to correct for that by cross-referencing against official complaint and recall data, not Reddit alone, but the underlying bias toward "things going wrong" is worth keeping in mind when you read a page.