Methodology

We make every reasonable effort to ensure that specifications published on this site are supported by an identifiable source. We do not treat patterns, automated matches, search snippets, or information copied from another site as sufficient evidence on their own.

Where a figure is recorded, we aim to attach the source, the source’s own words quoted verbatim, and where in the source the supporting information can be found.

A value without supporting evidence is not intended to be recorded as verified. A quotation that does not actually support the value it is attached to should not be accepted as evidence.

These requirements are built into the research and database tooling wherever possible. Nevertheless, mistakes can happen. If we have recorded something incorrectly, we want to know so that we can review the evidence and correct it.

A database that is 99% right is not 99% as useful as one that is right. A reader who finds one wrong figure has no way to know which of the others to believe, so one bad voltage discredits every voltage. That is why coverage grows slowly here: the alternative is a bigger catalogue that is worth less.

It is not a theoretical concern. One brand’s listing gave a battery as 15.63 Ah and 750 Wh; the brand’s own page said 21 Ah. The 15.63 had been produced by dividing the motor’s 750 W by the pack’s 48 V, and the 750 Wh was that same wattage relabelled — so multiplying the two back together returned 750 and every arithmetic check passed. It understated the battery by a quarter and only reading the page caught it.

Every compatibility row carries one of these, and they are ordered. Nothing is promoted because a specification happens to match.

• OEM — the manufacturer identifies this battery as the one for this exact model.
• Manufacturer documented — manufacturer documentation states the compatibility explicitly, without being the model’s own page.
• Vendor explicit — an authorised vendor or a marketplace listing states it explicitly. Never inferred from a seller merely stocking both items.
• Physical match — real specifications line up, and nothing states compatibility. This is not a recommendation.
• Unverified — there is not enough evidence. This is the honest floor.

Where this catalogue sits on that scale today: published rows sit on the top two rungs and the comparison rung, and nowhere lower. Most are manufacturer documented, which in practice means the BATTERY maker naming the cart. A few are OEM — the CART maker naming the pack their vehicle ships with, which is rarer and stronger, because it is the company that built the vehicle saying what goes in it. Advanced EV do that: they name an Eco Battery pack on several of their lines, and where they name one exactly, the row says OEM. Bolt Energy’s rows sit on the same rung as the other battery makers’: they make the packs, and their fitment guide and compatibility search name the carts, years and factory batteries each kit is built for, which is the pack maker’s own written statement rather than two specifications being compared. And some sit on the comparison rung, one step lower, where neither company names the other: a cart maker states what its cart carries — “(6) 8V Sealed Maintenance-Free AGM” — and a battery maker publishes a pack of that voltage, and we set the two side by side. Those rows say so in their tier, carry the words they rest on, and list what the comparison does not establish, which is usually everything except the voltage. Nothing is published from a rung below that. Where a row establishes that a pack will run a cart but not that it fits the battery bay, it says so on the row itself rather than leaving you to assume, and the whole scale is printed here so that if a weaker row ever appears you can see which rung it is on rather than taking a badge on trust.

A physical match is not compatibility. Two packs of the same voltage and capacity in the same case can differ in discharge limit and pin assignment — one manufacturer in this catalogue sells eight packs in one enclosure at two different voltages. Where we know packs share only a shape, the page says so and prints each one’s own figures beside it.

UL 2271 is a battery standard, ANSI/OPEI Z130.1 covers the golf car itself, and 49 CFR 571.500 is the federal standard for a low-speed vehicle. They are different claims about different things, and none is ever treated as evidence for another.

• Manufacturer claim — marketing text says “certified”, with no document or number.
• Document claim — a real manual, spec sheet or declaration of conformity states it. Note that a declaration of conformity is issued by the manufacturer about itself.
• Independent source — a document written by the certifying body, with a certificate number.
• Verified record — confirmed against the certifying body’s own register.

Nothing is promoted between these automatically. Most certification claims on the internet are the first tier, and we label them as such even when the manufacturer is almost certainly telling the truth.

Recall records come from the regulators themselves — the US CPSC, the UK’s OPSS, the EU Safety Gate, Product Safety Australia, Health Canada, Transport Canada and France’s RappelConso — through their own official data, never scraped from a summary site.

A regulator’s machine-readable feed is a summary of the notice, and the two do not always agree — one national feed lower-cases every word it returns, another has returned a headline with a word missing. Where they differ, the published notice is what we take, and it is the notice’s address that each recall page links.

Transport Canada’s notices are vehicle recalls, and the feed we receive them through gives neither the maker nor the models — both are on the notice itself, which is where we read them. Ten of the records we hold have been read and decided to be a golf car or low-speed vehicle, and the model lines they name are matched to the models here by a person reading each notice. The rest name vehicles this catalogue does not hold and stay unpublished: a record we hold and have not reviewed is not a record we publish.

Matching a recall to a product is done with a controlled confidence scale, and only exact and likely matches are shown publicly. A weaker match keeps its record internally rather than putting a hedged warning on a page it may not concern; hedging does not make a jacket-battery recall useful on a cart page.

A recall’s own scope travels with it. Most recalls cover particular serial numbers, colours or purchase dates, and where the notice says so the page shows that text and tells you to check your own unit. The absence of such a marker never means your unit is unaffected — only that the notice we read did not state a limit.

Where nothing matched, the page says no matching record was found in the available data for that jurisdiction. It never says the product is safe.

Manufacturers contradict themselves more often than you would expect, and the honest response is to publish the disagreement rather than quietly pick a side. One manual states a pack as “48V 13Ah, 768Wh” — but 48 × 13 is 624, the part number in the same line encodes 13 Ah, and two sibling manuals print 624 Wh for the identical part. That is recorded as the manufacturer’s error, with the evidence, and the catalogue keeps 624.

Where we genuinely cannot tell, nothing is changed and the conflict is recorded. Where a manufacturer’s own figure is simply unusual — several brands rate pack energy slightly above nominal volts × amp-hours — we keep their figure rather than replacing it with our arithmetic.

Manufacturers delete pages constantly — a model is discontinued, a site is rebuilt, a spec sheet is quietly corrected — and when that happens every quote taken from it becomes unverifiable. So we keep our own copy of the documents we cite, with the date we took it and a cryptographic hash of the bytes.

Each figure names its source beside it, written out in full as plain text rather than as a link. The address is the manufacturer’s own, so it can be opened directly and a fact checked against its origin, which is stronger than checking it against us. If that address no longer works, ask: where we hold a copy of what we read, we will supply it.

We do not use an AI model as a source of truth. Language models are used to extract text from documents and to flag contradictions; they never decide that something is compatible, safe or certified, and nothing reaches a page because a model proposed it.

We do not generate product images, invent part numbers, fabricate reviews or ratings, or publish a specification we could not attribute. We do not present ourselves as the seller of anything.

Every figure on a product page carries its source. Open it. If the source does not say what we say it says, that is a defect and we want to know — write to us through the contact page and it will be corrected.

Please read this alongside the Disclaimer, which explains the limits of everything above. Verify anything here against the manufacturer’s own documentation before you buy, fit or charge anything.