Fuel gets into engine oil. That much is measured, mechanically understood and not in dispute. What is worth arguing about is how much, in which engines, and whether the number on a used-oil report describes the oil that is actually in your bearings.
- Applies to
- Petrol direct injection, and diesels with a particulate filter
- Warm petrol engines
- Around 1-4% by mass in collated studies
- Diesel regeneration
- Over 20% temporary dilution reported
- Evidence
- Peer-reviewed review, SAE engine studies, ASTM test method
- Australian court record
- No oil-dilution finding in the Toyota DPF judgment
How fuel gets into the sump
There are two mechanisms, and they belong to two different kinds of engine.
In a petrol direct-injection engine, fuel is sprayed into the cylinder rather than into the inlet port, so it has less time and less hot surface to evaporate against. The 2021 review Fuel-Lubricant Interactions describes wall wetting caused by direct impingement of fuel sprays onto the cylinder wall, and states that excessive fuel dilution can wash lubricant films off piston cylinder walls. Fuel that has not evaporated lands on the liner and the piston crown, the oil film there absorbs it, and the rings scrape it down into the sump. Redesigning injector spray angle is a documented mitigation.
In a diesel with a particulate filter, the fuel is put there on purpose. To get the exhaust hot enough to burn the filter clean, the engine injects fuel late in the cycle that is not meant to burn in the cylinder at all. Wattrus, writing for Sasol Technology, states that these late injections do not atomise and evaporate as readily as during normal combustion, causing a portion of the fuel to impinge on the cylinder wall and wash down into the sump, which can degrade the oil and lead to increased engine wear and/or engine failure.
Neither is a defect. Both are the cost of a specification - inject in-cylinder for efficiency, burn the soot off downstream for emissions - with the fuel that ends up in the oil managed by a service interval rather than by margin in the design.
Why the percentage depends on where the sample was taken
In fully warmed-up, steady-running petrol engines, measured and modelled fuel dilution is typically low single-digit - around 1-4% by mass in the studies collated by Taylor - though those measurements were made on older port-injection engines without modern aftertreatment. Direct-injection engines are not covered by those figures: the same review notes wall wetting makes dilution more of an issue in gasoline direct injection, reports just over 3% in a hybrid turbocharged GDI vehicle against 1.5% in its conventional equivalent, and records over 20% temporary dilution in a direct-injection diesel during filter regeneration.
Two things then make the popular numbers look worse than the oil in your engine.
The first is temperature. The viscosity collapse quoted in most coverage is a 40 degree C figure: the same review reports that 20% dilution takes a typical SAE 5W-30 from around 55 cSt to 23 cSt at 40 degrees C, and that dilution levels of 10-20% measured at 40 degrees C are likely to be reduced to less than 1% at 150 degrees C. A cold-sample viscosity result therefore overstates the penalty on the oil film that is actually doing the work, for a petrol engine.
The second is that fuel leaves the oil as well as entering it. Tormos and colleagues measured a fuel change of minus 13.99% by infrared analysis at a conventional 1200 rpm, 2.5 bar operating point - direct measurement of net fuel removal during ordinary running. That is the physical basis for the standard advice that a decent sustained drive helps, and it is the rare piece of workshop folklore with a measurement behind it.
Two things that are widely repeated do not hold up. There is no industry condemnation limit for fuel in oil that we could source: ASTM D3525 is a gas-chromatographic measurement method, not a threshold, so any 2%, 3% or 5% figure should be attributed to the specific laboratory or manufacturer that set it. And a rising oil level on the dipstick is consistent with dilution but is not a measurement of it - the only quantification is a lab test on a sample.
Diesel is the case that persists
Petrol dilution largely boils back out. Diesel does not, because its boiling range sits above much of the sump temperature window. Taylor reports that significant levels of diesel fuel dilution could persist at oil temperatures in the range of 100-150 degrees C - the same window in which petrol dilution falls below 1%.
The measured numbers come from the regeneration points. Tormos and colleagues recorded an infrared fuel change of plus 15.07% at a 2000 rpm, 2 bar regeneration point, rising to plus 22.28% with delayed main injection and plus 32.93% with advanced post-injection, with kinematic viscosity at 40 degrees C falling 6-6.5%. Those are steady-state dynamometer points, two of them deliberately mistimed, and the figures are fuel change against a reference rather than absolute fuel content in the sump - the same work is summarised elsewhere as temporary dilution of over 20% during regeneration.
The duty cycle that drives it is an ordinary Australian one. Wattrus describes the late post-injections that regeneration requires, and the sump temperature effect is the one Taylor reports - diesel dilution persisting at 100-150 degrees C. Short suburban trips keep the sump at the cool end of that range, which is the end where the fuel is least likely to evaporate back out. If you drive a diesel ute or 4WD on school runs and hardware trips, that is the mechanism that applies to you.
Servicing a short-trip diesel on the headline interval instead of the manufacturer's severe-service schedule is the expensive version of this problem.
What it actually does to the oil
At high dilution the effect on the oil's ignition behaviour is dramatic. Hu and colleagues measured the flash point of an SAE 5W-30 falling from 245 degrees C to 90 degrees C with 15% fuel content, and the fire point from 265 degrees C to 150 degrees C, with low-speed pre-ignition frequency correlating strongly with the minimum auto-ignition temperature of the oil particles. Note the level: 15% is far above the warm-engine equilibrium, so that is the severe end of the range rather than normal service.
Dilution is also not the whole story on pre-ignition. Fuel properties and lubricant chemistry are both documented drivers and they interact rather than act separately: high-aromatic or high-distillation-temperature fuels increase activity, calcium sulfonate detergents tend to increase it while magnesium sulfonate detergents tend to decrease it, and a low-activity fuel was insensitive to the lubricant used while a high-activity fuel could be moderated by a low-activity lubricant, per Kocsis, Briggs and Anderson at Southwest Research Institute. That paper examines fuel composition and oil additive chemistry, not dilution of the oil by fuel.
Ethanol-blended fuel also ends up in the sump, with 6-25% ethanol and its combustion products reported across field and laboratory tests in a 2023 review, mostly under cold running, short trips, start-stop operation and higher blends. Bench work adds a concern not demonstrated for petrol dilution: at 5% ethanol and 70 degrees C the anti-wear tribofilm formed more slowly, ended up thinner and shifted chemically. The limits matter as much as the finding - the effect disappeared at 100 degrees C because the ethanol evaporated, it was worse with water-bearing ethanol, and no petrol-dilution control was run, so this reads as not shown for petrol rather than petrol is harmless.
What we could not find is the study everyone wants: no primary work measuring wear against dilution level in an engine specified for a low-viscosity oil was reached in this research. The mechanism is plausible and the measurement appears not to have been published.
The specification does not carry this either. API's published gasoline service categories describe API SP in terms of low-speed pre-ignition, timing chain wear, high-temperature piston and turbocharger deposits, and sludge and varnish control; fuel dilution is not among the properties API names. API does not publish per-test limits on that page, so that is a description of what the category is marketed as covering, not proof that no certification test touches dilution. In practice the service schedule carries the load: follow the interval and grade in your handbook, and shorten it if the car does mostly short, cold trips.
The evidence that cuts the other way
The strongest argument against treating this as a modern-engine story is that the biggest published figures are from before modern engines. Taylor records petrol sump-oil dilution of 10-20% in 1998 field trials where cars were driven up to 10 km from cold and no further (Kollman and colleagues), and 3-5% in summer rising to 8-11% in winter on a comparable severe short-trip cycle (Bergstra and colleagues). Those were deliberately extreme duty cycles on older, pre-direct-injection vehicles. Both trials were on pre-direct-injection cars, which does not establish that a level seen in a particular modern engine under normal use is acceptable, but does show that severe short-trip use produced double-digit dilution before direct injection existed.
The same review offers no direct comparison of direct-injection against port-injection dilution at the same duty cycle. The claim that one is measurably worse than the other is not established there.
And the short-trip framing itself is incomplete. In one detailed study of a highly boosted turbocharged direct-injection engine, Hu and colleagues at Jiangling Motors found dilution rising with high-speed rated torque, because longer injection durations put large liquid droplets onto the piston top, reaching a maximum of up to 9% in a cyclic-load test simulating customer driving. That is one engine from one manufacturer on a dynamometer schedule, not a fleet result, so do not read 9% as a figure for direct-injection engines generally. But it points the opposite way to the usual prescription: gentle driving is not a universal answer.
What the record says, and what it does not
Named engines are where care is owed. The verifiable record for Honda's 1.5-litre turbocharged engine is a United States federal class action in the District of Minnesota, docket 0:18-cv-01549, filed 4 June 2018 and terminated 11 September 2020, with a consolidated amended complaint filed 30 November 2018 concerning an alleged oil dilution defect. An allegation in a filed complaint is not a finding. A direct query of NHTSA's recall database for the 2018 CR-V returns five campaigns - fuel pump, electric power steering, seat belt buckle, a second fuel pump and a passenger weight sensor - and none concerns engine oil or fuel dilution. The response there was service bulletins and litigation, not a safety recall.
Australia's landmark diesel case is routinely misquoted on this point. In Williams v Toyota Motor Corporation Australia Limited (Initial Trial) [2022] FCA 344, decided by Lee J on 7 April 2022 and covering 264,170 HiLux, Prado and Fortuner vehicles with 1GD-FTV and 2GD-FTV engines, the established consequences of the particulate filter defect were excessive white smoke, foul-smelling exhaust, filter blockages, excessive notifications and increased fuel consumption. Engine oil dilution does not appear among the findings.
We could not complete a search of the Australian recall register in this research, so we make no claim either way about whether any Australian recall for oil dilution exists. We also could not find a primary study isolating idling as a variable, despite how often it is named - the documented variables are cold start, trip length, ambient temperature, injection duration and post-injection timing.