Best Brokers

Editorial policy

Editorial policy

The rules that govern what gets published here, who signs it, and what has to be true before a page goes live.

Sourcing

We rank sources in this order, and we always say which one a figure came from:

  1. The provider’s own published documents — fee schedules, product pages, terms, disclosures. Almost every number on this site comes from here.
  2. Regulators and self-regulatory organizations — the SEC, FINRA including BrokerCheck, and SIPC. Every legal entity, registration status and enforcement matter comes from here.
  3. Established financial media — used only to corroborate a figure or to locate a primary document, never as the sole basis for something we publish.

We do not use competitor comparison sites as a source. If another comparison site is the only place a figure appears, that figure does not appear on ours.

Verification

  • Every published figure carries a verification status and a source reference in the data layer — this is enforced by the type system, not by discipline
  • A figure we could not read renders as “not verified” — never blank, never estimated. A blank cell reads as zero, and that would be a lie by omission
  • Every page shows the date its data was read, and we do not refresh that date without re-reading the source
  • A broker whose fee schedule we cannot read is excluded from every ranking, with the reason published on its own page

Claims about testing

We have not opened an account, deposited funds, placed a trade, contacted customer support or measured execution quality at any broker on this site. We write nothing that implies otherwise, and the five scoring categories that would require that work carry zero weight.

If that changes, the change will be described specifically: what was tested, on which accounts, on which devices, and when. A general claim of “hands-on testing” without those particulars is not something we will publish, and it is not something you should accept from anyone else either.

Bylines and review

Every ranking, review and guide carries a named author. Until a named person is assigned, the page displays an explicit notice where a byline would be, and no Person structured data is emitted. This is enforced in code: a placeholder author record cannot render as a name.

This is a single-operator publication. There is no second reviewing editor, and we do not display a “reviewed by” line, because one person cannot review their own work and claiming otherwise would be the kind of unfalsifiable process claim this site exists to avoid. What stands in for a second pair of eyes is mechanical rather than human, and it is public: every figure is wrapped in a structure that requires a source and a date, unverified fields cannot render as values, scores are recomputed from the data at every build, and an automated audit checks all 66 pages against nine invariants — including one that fails the build if any page claims testing that did not happen.

If a second named reviewer is added, this section will say so. A named byline commits its holder to having:

  • Checked every cited figure against the source cited
  • Confirmed that each scoring rationale matches the evidence behind it
  • Confirmed that nothing on the page claims testing that did not happen

We publish only credentials a reader can independently verify. We will not describe anyone as an expert without saying what makes them one and how you could check it.

Editorial and commercial separation

This is a mechanism rather than an assurance. Commercial relationships live in a single module that the scoring engine does not import; there is no code path from a commercial relationship to a score. Outbound links are constructed by one function, so a sponsored link cannot appear without the rel attribute and the row-level disclosure that go with it.

Ranking order is computed at build time from published sub-scores and published weights. Ties break on safety, then on cost. No commercial input exists at any point in that calculation, and you can verify that by recomputing any total yourself.

Use of automation

Research and drafting on this site are assisted by software, including large language models. That does not soften the sourcing rule: a figure is published only if it was read from a primary source, with that source recorded alongside it. No figure is published because a model produced it. Where automation could not reach a source — a page that blocked it, a document it could not parse — the field is marked unverified rather than filled in from the model’s own recollection. Several fields across this site are marked unverified for exactly that reason.

Publishing thresholds

A ranking is published only when every pillar it weights is scored for every eligible broker. A ranking that depends on an unscored pillar is built, labelled a draft, excluded from search engines and carries no table. We would rather publish the reasoning behind an unfinished ranking than a finished-looking ranking we cannot defend.

Financial content standards

  • Nothing here is personalized investment advice, and no page implies that it is
  • No projection, guarantee or implication of future returns appears anywhere on this site
  • Risk is disclosed on every page carrying a ranking, review or comparison
  • Promotions and sign-up bonuses are excluded from scoring entirely
  • Where a topic requires licensed advice — tax treatment, suitability — we say so and stop

See also: methodology, corrections policy, affiliate disclosure.