Editorial standards
How this magazine is made, and how to check it
KeepPickling uses an automated editorial desk working in the voices of eight editorial personas, under human editorial direction. This page explains exactly what that means, what the evidence labels on every story are for, what the desk is forbidden to do, and how to get something corrected.
Last updated 10 November 2026 Applies to every story
The short version
- The writers are personas. They are not hidden staff or real named people pretending to be something else. Each has a documented life and a written voice contract.
- The writing is AI-assisted. Drafts are produced by a model prompted against those contracts, then checked by automated quality gates. A human editor sets the standards, owns the queue, and handles corrections.
- Every story is labelled by evidence. The label says how the piece knows what it knows. See the five labels below.
- Fabrication is a firing offence. The desk may not invent tests, studies, prices, quotes, interviews, or human faces. Corrections are free and welcome at [email protected].
Who writes here
KeepPickling launched with a roster of eight writers. You can read each profile on the writers page. Each one has a role, a home, an age band, a playing level, the sections they cover, and a short list of facts they are allowed to draw on.
They are editorial personas. That is a deliberate choice for a small independent magazine that publishes every day: it keeps a clear, consistent first-person voice for each section without inventing a fake staff of photographers and correspondents. The personas are not presented as real people you can meet, and their faces are monogram marks rather than AI-generated portraits.
A persona is a constraint, not a licence. The desk may write about the documented life and the game around it. It may not invent a house, a diagnosis, a tournament result, a spouse's career, or a hobby that is not in that writer's documented facts. When a story needs something outside the record, the detail is left out rather than made up.
How a story is made
- A human editor sets the brief. The topic comes from a written editorial queue with a section, an angle, an evidence plan, and a date. It is not whatever a model feels like writing that morning.
- The desk assigns a writer who fits. Only a persona whose beats cover the section, and whose documented life supports the angle, gets the piece.
- The model drafts. It works from the persona's voice contract and a hard list of rules: no fabricated testing, no invented sources, no fake quotes, no AI marketing filler, no em dashes, at least two internal links to real stories, a clear takeaway near the top.
- Automated gates check the draft. The piece is rejected or repaired if it claims testing that was not done, cites a source that cannot be verified, contains banned filler, breaks the structure rules, links to a page that does not exist, or sounds nothing like the assigned writer.
- A human editor owns the standard. The editor writes and revises the rules, sets the queue, samples published work, reviews flagged pieces, and answers corrections. In autonomous mode the desk can publish a piece that passes the gates without a human reading every line first. That is a real trade-off: the magazine can publish daily, and the labels, the gates, and this corrections process are the safeguards.
- Publication is labelled. Every story shows its evidence label, a byline that links to the writer's documented profile, and an illustration that is labelled as an illustration. Hero images are generated, not photographs of a real court visit, and they never use AI-generated human faces to stand in for a writer or a reader.
The five evidence labels
Each story carries one label, visible near the byline. The label describes what the piece rests on, not how good it is. A judgment piece is not worse than a sourced piece, and a first-person piece is not a lab report. Two labels print under a friendlier name on the story page: experience shows as "First person", and derived shows as "Analysis". The categories below are the underlying ones.
| Label | What it means | What it does not mean |
|---|---|---|
| Experience (shown as "First person") | A first-person piece written inside the persona's documented history. The writer tells it from their own side of the net. | A report of a real named person's field test. It is not evidence of measured performance and it is not used to claim a product was tested against another. |
| Reported | Built on actual reporting by the human editor: interviews, notes from a court or event, documents obtained directly. The desk cannot earn this label on its own. | Something a model wrote in the style of reporting. If no human did the reporting, the label is not available. |
| Sourced | Factual claims link to a named, public source that a reader can check: a governing body rulebook, a manufacturer specification sheet, a published study, official health guidance. | That KeepPickling tested anything. A source is not a substitute for our own measurements, and we say so when a reader might assume otherwise. |
| Derived (shown as "Analysis") | A conclusion calculated or assembled from KeepPickling's own archive or from simple public data, with the working shown where it matters. | Original research on a population, a clinical trial, or a survey we did not run. |
| Editorial | Judgment, opinion, taste, and reflection. The piece is making an argument and should read like one. | An excuse for made-up facts. A factual sentence inside an opinion piece still needs to be true or sourced. |
Experience (shown as "First person")
What it means. A first-person piece written inside the persona's documented history. The writer tells it from their own side of the net.
What it does not mean. A report of a real named person's field test. It is not evidence of measured performance and it is not used to claim a product was tested against another.
Reported
What it means. Built on actual reporting by the human editor: interviews, notes from a court or event, documents obtained directly. The desk cannot earn this label on its own.
What it does not mean. Something a model wrote in the style of reporting. If no human did the reporting, the label is not available.
Sourced
What it means. Factual claims link to a named, public source that a reader can check: a governing body rulebook, a manufacturer specification sheet, a published study, official health guidance.
What it does not mean. That KeepPickling tested anything. A source is not a substitute for our own measurements, and we say so when a reader might assume otherwise.
Derived (shown as "Analysis")
What it means. A conclusion calculated or assembled from KeepPickling's own archive or from simple public data, with the working shown where it matters.
What it does not mean. Original research on a population, a clinical trial, or a survey we did not run.
Editorial
What it means. Judgment, opinion, taste, and reflection. The piece is making an argument and should read like one.
What it does not mean. An excuse for made-up facts. A factual sentence inside an opinion piece still needs to be true or sourced.
If a label and the story seem to disagree, that is a correction worth sending. It is exactly the kind of thing this page exists to make checkable.
What the desk may not do
These are hard rules, written into the prompt and into the automated gates. Breaking one is a defect, not a style preference.
- Invent tests or measurements. No "after eight weeks", no "our testers found", no implied comparison trial, no numbers from a session that did not happen.
- Invent studies, statistics, or expert opinion. No citation to a study that does not exist, no statistic without a checkable source, no quote attributed to a clinician or coach who was never asked.
- Invent prices, stock, specs, or model names. If a price or specification is stated, it is dated and checked against a real listing at the time of writing, and it is not presented as permanent.
- Invent interviews, reader letters, or conversations. Illustrative dialogue inside a first-person story is clearly the writer's memory, not a fabricated third party.
- Use fake human faces. No AI portrait passed off as a writer, a source, or a person at a court. Writers are represented by monograms and stories by labelled illustrations.
- Hide the automation. We do not present AI-assisted drafting as fully human reporting, and we do not bury this page.
- Use fake urgency or scarcity. No countdown timers, no "$X left", no "deal ends tonight" unless a live retailer listing proves it and the date is shown.
- Take money for coverage or ranking. No sponsored posts, no pay-to-play placements, no ranking by commission. Affiliate links, when they exist, are disclosed and never decide what gets covered.
- Give medical advice. The body and health section is practical, sourced writing about playing for a long time. It does not diagnose, treat, or tell a reader to stop or start a treatment.
- Pretend to have been somewhere. Travel writing says what is documented firsthand and what is researched from public information. We do not fake a site visit.
What the desk can do instead: link the source, show the working, mark the claim as judgment, say "I don't know", or leave the detail out. A shorter honest piece beats a padded confident one.
Corrections
Being automated does not make us unaccountable. It makes corrections more important, because a model can produce a confident sentence that is wrong. Email [email protected] with the subject "Correction" and include:
- the page address and the passage you are querying;
- what you believe is wrong and why;
- a source or a link if you have one, which speeds everything up.
What happens next. A human editor reviews the piece against the source. If we got it wrong, we correct it. Substantive factual errors get a dated correction note at the foot of the piece, saying what changed, rather than a silent edit. Small typo and style fixes do not need a note, but they are logged. If a claim cannot be supported, we cut it, rewrite the passage, or withdraw the piece with a note explaining why. If you asked, we tell you the outcome.
We aim to acknowledge a correction within five working days. A fact check that needs a real source can take longer, and we will say so rather than go quiet. Corrections are the same process whether a piece was drafted by the desk or written by the human editor.
Disagreement is not the same as error. If you think a judgment call is wrong, say so; we may add context, publish a letter, or simply disagree. If a fact is wrong, we fix it.
Independence and money
KeepPickling does not sell coverage. No brand can buy a review, a place in a list, or a kinder score. There are no affiliate links live on the site today, and the affiliate disclosure explains what will change when the shop opens. If a story ever contains a link that can earn a commission, it is disclosed at the top of the piece and again where the link appears.
The writer profiles show the documented facts behind each byline, the stories show their evidence label, and the sources inside a piece are there to be clicked. That is how a reader checks our work instead of taking our word for it.
Questions
If something here is unclear, ask. Write to [email protected] or use the contact page. Standards questions, corrections, and story tips all reach the same small desk.