Our editorial standard is one sentence: the review reflects the measured product, not the payer. This page states how that is enforced, who is accountable, and what happens when we get it wrong.
What is SirenPicks' editorial policy?
Scores come from a published five-criterion protocol and nothing else. Brands cannot buy a score, a ranking, a placement or inclusion in a roundup, and there is no sponsored content or advertorial anywhere on this site. Every review carries a named accountable author, samples sent for review are disclosed on the article, and corrections are published and dated rather than quietly applied. The one standing conflict — we own the SirenPicks device line — is disclosed on every page, labelled on every pick and handled identically at the bench.
- No pay-to-rank, at any price. There is no sponsored slot to buy and no advertorial to commission. A device appears because it cleared the protocol.
- Material safety is a gate, not a weighted average. An unverifiable skin-contact material excludes a device entirely.
- Accountability is a person, not a masthead. Mara Ellison owns device verdicts; Dr. Lena Ortiz owns body-safety claims.
- Corrections are public. Acknowledged within two working days, resolved within ten, with retraction when a review cannot be supported.
No pay-to-rank
Brands, distributors and retailers cannot buy a score, a placement, a rank in a roundup, or inclusion in a face-off. There is no sponsored content on this site, no advertorial, and no "partner pick" slot. A device appears because it cleared the protocol, and it is removed when it stops being the right recommendation.
Samples sent for review are disclosed on the article itself. A sample is a convenience, not a favour: it does not change the method, the weights or the verdict, and a device sent for free can — and does — receive a negative score.
What that means in practice
Policy statements are easy to write and easy to dodge. Here is the same policy expressed as the decisions it produces — the situations we are actually asked about.
| The situation | What we do | What we will not do |
|---|---|---|
| A brand offers payment for a positive review | We decline and, where the offer implies a policy breach, we stop engaging with that contact. | Take the money, or soften the eventual review to keep the relationship. |
| A brand sends a free unit | We disclose it on the article, test it on the identical protocol, and publish whatever it scores. | Treat a supplied unit as bought goodwill. Several have scored poorly. |
| A brand asks to approve the copy before publication | We offer a factual-accuracy window limited to specifications and claims about its own product. | Allow approval, veto or rewrite of a score, a verdict or a conclusion. |
| A brand threatens to withdraw future business | We publish as measured and note the attempt if it is material to the reader. | Delay, soften or spike a review for commercial reasons. |
| We discover a measurement was wrong | Correct it in place, date the change, and explain what changed. | Quietly edit the number and hope nobody checks. |
| A review cannot be supported by its measurements | Retract it, with the reason stated on the page. | Leave an unsupportable verdict live because it ranks well. |
| A competitor outperforms our own device | Rank the competitor first and state the criterion it won on. | Bury a rival result because we own our own line. |
| A brand disputes a score publicly | Publish the underlying measurements and the weights, and the editor's ruling on the dispute. | Respond by negotiating the score in private. |
Ownership and conflict of interest
SirenPicks owns the SirenPicks device line, and SirenPicks products appear in our own roundups and face-offs. This is a real conflict of interest and we treat it as one rather than hiding it:
- Disclosed everywhere. The ownership statement appears in the footer of every page, on this page, in the Trust Centre, and in the machine-readable organisation data.
- Labelled on the pick. Every SirenPicks entry in a roundup or face-off carries an explicit "ours" marker. No reader should ever have to check a company registry to find out who made a recommendation.
- Identical protocol. SirenPicks devices run the same five tests, on the same equipment, with the same published weights as every competitor. No exceptions, no "internal" criteria.
- They lose sometimes. Where a competitor outperforms a SirenPicks device on a weighted criterion, the competitor is ranked first and the reason is stated in the verdict.
- You can opt out entirely. Ask us for the best non-SirenPicks device for your constraint and we will name it. That recommendation is editorial, not commercial.
If you believe this conflict makes our recommendations unusable for you, that is a legitimate position, and our reviews are written so the underlying measurements are published either way — read the numbers and the verdict will still be useful.
Who is accountable
Every review, roundup and face-off carries a named author. Mara Ellison is accountable for device testing and verdicts; Dr. Lena Ortiz is accountable for material verification and body-safety claims. There is no anonymous house byline on this site. If a claim cannot be attributed to a named person, it does not get published.
Sources, data and AI
Numbers published on SirenPicks come from our own bench or from a manufacturer spec sheet, and the review states which. We do not invent a measurement to fill a gap: if a figure has not been verified, the review says "unverified" rather than quoting the box.
What a claim has to rest on depends on what the claim is, so the rule runs per claim rather than per page. The chart below sets out the five kinds we publish and the one form of proof each accepts. Material safety is the strictest because it is the only one where the reader cannot check the answer themselves and the damage is hardest to undo, which is why it takes a lab report naming the standard and not a statement from the brand. Noise, runtime and dimensions are ours to measure on a protocol we publish.
Market size is quoted from a named, dated report. Factory price and MOQ are the factory's figures, dated, and never rounded in a direction that flatters the offer. The ordering is the same logic a reader would apply unaided — it is the idea of a hierarchy of evidence in the Wikipedia entry of that name, applied to a shopping category instead of a clinical one. The last row is the one that makes the rest usable: when the proof a claim requires is missing, the page says so in those words.
We use software, including language models, as drafting and research tools. We do not publish machine-generated text as a review, and no score, verdict or measurement is ever machine-generated: a person runs the test and a named person signs the verdict. Pages are written for a human buyer first — the structured data on them exists so machines can quote the page accurately, not so machines can write it.
We also keep our own optimisation inside Google's published rules. When Google extended its spam policies to AI Overviews and AI Mode on 15 May 2026 — and reinforced it through the June and August 2026 spam updates — it made one thing explicit: internal linking that exists to signal AI-targeted content, rather than to serve a reader, is spam. We treat every internal link as a reader task, each one resolving to the one page that answers a specific question, and we publish a direct answer at the top of each section so an answer engine can quote it without our having to manufacture links. The official changelog is on Google Search Central — Documentation Updates.
Each row accepts one kind of proof and no other: when the proof above is missing, the claim drops a level instead of staying where it was.
What each kind of claim has to rest on
The top row is the only one that demands a third-party document; the rest are our own measurements or dated figures from someone named.
Corrections and complaints
When we get something wrong we correct it in place, keep the original wording visible where it materially changes the verdict, and date the change. We do not silently delete an error.
The clock is written down, not implied: a missing or unclear disclosure is fixed the same day, a factual error is acknowledged within two working days and resolved within ten, and a disputed score is answered with the measured data and the weights rather than with an opinion. The trust page sets that clock out as a chart so a reader has something to hold us to instead of a promise.
- Report an error: contact the editorial desk. Tell us the page and what is wrong.
- Response: we aim to acknowledge every correction request within two working days, and to publish or explain the outcome within ten.
- Escalation: if a correction is disputed, the editor's decision is published alongside the correction rather than buried.
- Retractions: if a review cannot be supported by the measurements, the review is retracted and the page says so.
Independence of the commercial side
The partnerships function — testing submissions, data licensing and retail sourcing described on the For Brands page — has no ability to edit, delay, soften or withdraw a review. Commercial enquiries are answered by the partnerships desk; verdicts are written by testers. Nobody who receives commission can change a score.
Privacy and reader data
We keep reader data collection to the minimum needed to answer you and to see which tests are read. Analytics is optional, off until you allow it, and never used for advertising or cross-site tracking. Full detail, including every cookie and your rights, is in the privacy policy.
Hold us to it
This page is only worth reading if it is enforceable, so here is how to enforce it. If you see a sponsored-looking placement, an undisclosed sample, an unattributed claim, an undated correction or a score that does not reconcile with the published weights, send it to the editorial desk. Correction requests are prioritised over every other kind of message.
How much of the site rests on a supplier's word
The right-hand bar is not a defect: a wholesale catalogue cannot measure a unit nobody has built yet. What it can do, and does, is say so in the caption of every table.