How VoltRipper verifies a spec — and what we still can't tell you

It means a human-readable claim on this site traces back to a source we can name and you can open.

Every bike in the catalog carries its sources with it. Right now that is 137 source entries across 52 bikes, and zero bikes with no source at all. When a manufacturer publishes less than we would like, the record says so rather than filling the gap with an estimate that looks like a fact.

It also means something narrower than most review sites imply. We have not ridden these machines. We do not own a dynamometer. When you read a top speed here, you are reading what the manufacturer published, what a dealer listed, or what an independent outlet measured - attributed to whichever it was. You are not reading our stopwatch.

• No hands-on testing. Not "extensive testing," not "we found," not "in our experience." If you find that language anywhere on this site, it is a defect - tell us and we will fix it and log it.
• No measured performance figures of our own. Our range estimates are explicitly modelled, and the model shows its arithmetic.
• No generated image that impersonates a photograph. Illustrations explain concepts. Product photography is real product photography, and it is labelled.

Manufacturer range claims in this category are marketing numbers, and the gap is consistent enough to measure. We track both figures wherever both exist - currently 31 bikes with a published claim and a credible real-world figure.

Across that set, real range averages 59.4% of the claimed figure.

That ratio is not a rule of thumb we inherited. It is computed from our own catalog, it is recomputed as the catalog changes, and it is the coefficient behind our range calculator. Only a handful of the bikes in the set sit near the ratio itself, so it is not a circular result - it is a real distribution with real spread, and the calculator outputs a band rather than a false-precision single number.

Alongside it we keep 103 reality checks - individual claim-versus-observed comparisons on specific bikes. Every one of them carries a source link. Where the only available counterweight to a manufacturer claim is the manufacturer's own fine print, the check says that too, because "the maker quietly caveats this themselves" is weaker evidence than an independent measurement and you deserve to know which you are getting.

The 50-state legality set is sourced separately and more strictly, because people act on it.

It currently rests on 298 source entries, of which 242 are unique .gov URLs - state legislatures, DMVs, DOT and OHV authorities. Each state page lists what it is built on and when it was last checked.

We are candid about its uneven quality. An August 2026 audit of our own data found that coverage is inverted: several of the largest states are among the thinnest, some records lean on secondary sources we would not choose today, and a commercial registration service appeared as a citation on pages where a statute belongs. That is being repaired state by state, largest first, using primary sources only. Until a state's page cites its controlling statute by section number, treat it as a starting point for your own check with your DMV - not as the last word.

Nothing on this site is legal advice.

Google asks publishers to make automation self-evident to readers. Here is ours, plainly.

VoltRipper is built by AI agents working under human direction and human accountability. Research, drafting, data verification and code are performed by AI systems. A human owner sets the direction, approves what ships, and is answerable for it.

We think that is a strength worth stating rather than a fact worth burying, for one reason: it makes the method inspectable in a way most review sites are not. The trade-off is real too - an agent can verify a number faster than a person and can also be confidently wrong in ways a rider never would be. Everything in the next two sections exists because of that second half.

What automation does not do here: it does not invent specs, it does not generate a review of a bike that is not in the catalog with sources attached, it does not decide rankings by commission, and it does not produce an image that pretends to be a photograph.

Because the failure mode of an automated pipeline is a confident wrong number, the numbers are checked by machine on every build. Not as a formality - these run in CI and a failure blocks the deploy.

20 automated checks currently gate every release, backed by 60 test files. Among them:

• Score consistency - every Score stated in prose must match the Score the model computes. This caught four pages arguing from a Score relationship that no longer existed.
• Price consistency - every price in prose must match the catalog. This caught five inverted price claims and a stated gap that was $2,000 wrong.
• Catalog claims - counts and superlatives written in prose ("the only bike under $2,000 with...") are re-derived from the data and must still be true.
• Spec deltas - a spec that moves in the catalog flags every page that discussed the old value.
• Freshness - a page's "updated" date is compared against its actual git history. If we edited a page and did not update its date, the build fails. This one caught 230 files advertising a stale date, which had made a month of corrections invisible to search engines.
• Evidence language - prose is scanned for claims stronger than the sourcing supports.
• Internal links and images - no broken links, no unpublished targets, no bike without a verified photo.

These guards are the reason we can tell you the numbers above are current, rather than hoping so.

When we get something wrong, we fix it on the page, date it, and say what changed. We do not silently edit.

A correction here looks like this, taken verbatim from a live review:

Top-speed correction, July 2026: we previously listed this bike at 75 mph. Rendering Sur-Ron's own spec page directly, the published number is 68.3 mph - exactly 110 km/h, where 75 mph is exactly 120 km/h. The error was a unit conversion, not a source disagreement.

And from a comparison guide:

Correction (2026-07-31): this page previously said all three bikes tied at a Score of 86. They do not tie - they score 92, 91 and 90. The picks below have not changed, because they never rested on the tie; the reasoning did, and it has been rewritten.

That second one is the standard we hold ourselves to: say whether the conclusion moved, not just the number. A correction that quietly leaves the reader wondering if the recommendation changed is half a correction.

The public, citable record is the corrections log, with each correction anchored to its own entry.

Found something wrong? Tell us - corrections are the cheapest thing we buy. Until a contact route exists, the site's current ownership and affiliate-disclosure policy lives at /disclosure.

An honest list of this site's blind spots:

• Ride feel, ergonomics under load, and suspension quality. These need a rider. We report geometry and componentry; we cannot tell you what the bike feels like at speed on a rutted trail.
• Long-term durability. We can report warranty terms and known failure patterns from owner reports and press. We have not run a bike for two years.
• Real-world charge times and thermal behaviour under hard use, beyond what manufacturers and independent tests publish.
• Dealer service quality, which varies by location in ways no spec sheet captures.

Where any of these decides your purchase, a forum full of owners will serve you better than we can, and we would rather say so than pretend otherwise.