This page explains how Plainly Now decides what to publish, how an article is actually built, where automation is involved, what we do when we get something wrong, and who pays for any of it. It is long because the method is the point — you should not have to take it on faith.
Our promise
Four commitments govern everything below.
- Facts are tied to the record. Our workflow is designed to ensure that material factual claims are supported by identifiable sources recorded for the story — not offered as a hedge, not as a guess, not as “reports suggest.”
- Uncertainty is reported, not smoothed over. When a detail is unconfirmed, or when nobody actually knows yet, the article says so.
- Plain language is a duty, not a style. An explanation the reader cannot follow is a failed explanation, however accurate it is.
- Mistakes get fixed in public. We correct confirmed errors as quickly as one person reasonably can, and we do not quietly rewrite history.
Health and medical articles here are information, not medical advice — see our Terms of Service for the full disclaimer. In the United States, call 911 in an emergency, or call or text 988 if you are in crisis; elsewhere, contact your local emergency or crisis service.
How a story is made
Plainly Now is an AI-assisted publication. Here is the honest version of the process, in order.
1. A topic is selected
We start from stories a general audience is already asking about. We are not trying to break news first; we are trying to be the piece that makes a story make sense. The site publishes news explainers, not opinion or editorial content; if that ever changes, such content will be clearly labeled as opinion. We currently publish a limited number of individually processed stories per day, each one taken through the full review pipeline.
2. Research is gathered first
Before a word of the article is written, information about the topic is collected from multiple public sources into a single research record — an internal dossier of what those sources say. This record, not the writer’s memory or general knowledge, is the raw material for the article.
3. A draft is written from that record
The draft is generated from the research record. It is meant to be an account of what the sources support, not an essay about the topic in general.
4. Automated review checks the draft against the record
The draft then passes through multiple review stages. These stages check each claim in the draft against the research record, simplify language that has drifted into jargon or padding, and remove or flag statements that cannot be traced to a source.
5. A final correction stage patches what review found
Defects identified in review — an unsupported detail, a number that does not match the source, an overstatement — are corrected before publication rather than published with a note attached.
6. Images are selected under the rules below
See Images. The image rules are written to prefer no picture at all over one we cannot account for.
Sourcing standards
Multiple sources
A story is researched from more than one public source. We prefer original reporting, primary documents, official statements, and named on-the-record accounts over aggregation and over accounts attributed to nobody in particular.
Our own words
Articles are written from the research record in our own words. Verbatim reuse of another outlet’s expression, beyond a short quotation with attribution, is prohibited.
Tying claims to the record
Our workflow is designed to ensure that material factual claims are supported by identifiable sources recorded for the story. This is the single most important check in the process, and a claim that review cannot trace to the record is flagged, rewritten, or removed before publication.
Right of reply
When the research record contains a response or denial from the person or organization a story is about, the article reports it. Our automated process does not currently contact story subjects for comment before publication — a limitation we state plainly. Anyone covered by a story can respond at any time via contact@plainlynow.com, and substantive responses are added to the article.
When sources disagree
Most disagreements between sources are minor: a differing figure, a differing time of day, two descriptions of the same event that do not quite line up. In those cases the writer settles the question by choosing what the record best supports and writing it plainly, rather than handing you a list of contradictions to sort out yourself. When the disagreement is itself part of the story — when who says what is the news, or when a central fact is genuinely contested — the article reports the disagreement openly and, where possible, says who says what. Silence is treated the same way: if a question has no answer yet, the article says the question is open.
Rumors, allegations, and unconfirmed claims
We do not repeat a claim about a person’s health, finances, family, or conduct just because it is circulating. An allegation is described as an allegation, attributed to whoever actually made it, and paired with any response on the record. Sensitive subjects — illness, death, legal proceedings, matters involving children — carry specific additional requirements: each claim must be attributed to the source that made it, the article does not speculate about anyone’s health, diagnosis, or legal status, and the language stays neutral rather than dramatic.
Simplifying is allowed; changing the facts is not
Rounding a figure, dropping a technical qualifier that does not affect the meaning, or describing a mechanism in everyday words is part of the job. Altering what happened, implying a certainty the sources do not support, or inventing a detail to make a sentence flow is not, and is treated as a defect to be corrected.
AI transparency
We think readers deserve a specific answer to “how much of this was written by a machine,” so here it is, without hype in either direction.
What the automation does
- Gathers research from public sources into the research record.
- Writes the draft from that record.
- Checks each claim in the draft against the record, and flags or removes what it cannot trace.
- Simplifies language and structure toward plain English.
- Applies factual corrections identified in review before publication.
- Selects and labels images under the rules in the next section.
What the automation does not do
- It does not decide the standards. What counts as an acceptable source, what may be published about sensitive subjects, how images may be used, and what corrections are owed are editorial rules set by the publisher of this site, not preferences invented by a model.
- It does not override those rules. When the rules and a generated draft conflict, the draft loses.
- It is not a substitute for responsibility. The individual publisher of Plainly Now is responsible for everything published here, whatever produced the first draft.
Does a person read an article before it is published?
No. Publication is automated. The publisher reviews newly published articles soon after they appear and corrects issues after publication rather than gatekeeping each article before it appears. That is the design, and we state it plainly so you can weigh it yourself. Every published article also carries an AI-generated notice in its byline, where you actually read it. It reads “AI-generated article”.
What we are not claiming
We are not claiming that a human being rewrites every sentence — that would be untrue, and a trust page built on a comfortable untruth is worse than no trust page. We are also not going to pretend that “AI” makes the work either magical or worthless. It is a production method with specific strengths (consistency, tracing claims against a record, plain rewriting) and specific weaknesses (confident wording around thin evidence, and the fact that a checker can share a writer’s blind spot). Three parts of the process are aimed at exactly that weakness: a draft is reviewed independently by several different AI models rather than by one, claims that review cannot trace to the research record are flagged and removed, and a final correction pass looks specifically for confidence the evidence does not support. The rules above exist because of the weaknesses.
Accuracy, and the limits of this process
The process described here is built to catch errors before publication, and it does catch them. It is not perfect. Mistakes can still happen, and we do not claim otherwise. A review stage can miss an error, a source can be wrong, and a story can move on after we publish. What we can promise is the response: when an error is confirmed, we fix it, and we do not argue with a reader who was right.
Corrections
How to report an error
Email contact@plainlynow.com with the article title or link and a short description of what you believe is wrong. If you can point to a source, that speeds things up considerably. You do not need to be the subject of the article to report an error, and reports are reviewed regardless of tone.
What happens next
Corrections are handled by the publisher personally, not by an automated system: a person reads the email and checks the claim against the story’s research record, the underlying sources, any material you provide, and any newer reliable information — and takes appropriate corrective action when the report is substantiated, as quickly as one person reasonably can. Reports involving an immediate safety, privacy, or identification concern are prioritized. We review all correction reports, though we may not be able to send an individual reply to every general message or tip. If the claim is right but poorly worded or easy to misread, we rewrite it. If we conclude the article was accurate, we will say so and explain why rather than ignoring the message. Once a claim of inaccuracy or a legal notice is received, the underlying records for that story are preserved.
Five things we can do, and what each one means
Not every change to a published article is a correction, and calling them all one thing hides what actually happened. We use five plain labels.
- Update — new events have moved the story on. Nothing published was wrong; there is simply more to say.
- Clarification — what we published was accurate but unclear, or easy to read the wrong way. We rewrite it so it reads the way it was meant.
- Correction — we published something wrong. We fix it in the article and disclose the fix there.
- Editor’s note — a serious dispute has been raised, or a review is underway. We say so on the article while we work it out, rather than leaving readers to guess.
- Retraction — the core of the story turns out to have been unreliable. We withdraw it and say that we did.
How corrections appear
A substantive factual correction is made in the article and disclosed there, so that a reader can see what changed and when. We do not silently swap a wrong fact for a right one and pretend the wrong one never ran. Fixes to typos, broken links, and formatting are made without a note, because nobody is served by a correction notice about a misplaced comma. When an error was serious enough that people may have acted on it, we say so plainly at the top of the piece.
Removal requests
Accuracy alone does not necessarily require permanent publication in every form. We may correct, update, anonymize, deindex, restrict, or remove content when required by law or when safety, privacy, fairness, or the public interest warrants it. We do not remove accurate reporting merely because its subject dislikes it. If you believe an article is inaccurate, tell us what is inaccurate and we will look at the facts.
Images
Pictures make claims too, so they follow rules as strict as the text.
Real photographs
- We use real photographs only under licenses that permit publication: public domain or CC0, CC BY, or CC BY-SA images from Wikimedia Commons, and images under the Pexels license.
- We credit licensed images as their licenses require, and we correct any attribution errors we find or that are reported to us. For real photographs we also state the year the photograph is from when the source provides one, so a picture from years ago is not mistaken for one from this week.
- Place-and-time honesty. A caption states where and when a photograph was taken, or it claims neither. We would rather give you a bare credit than a plausible-sounding caption we cannot support. A photograph of a person is never presented as if it were taken at an event it was not taken at.
- Publishing a photograph of the wrong person is a worse failure than publishing no photograph at all, and the process is deliberately built to prefer no image over an uncertain one.
- For every real photograph we keep an internal record of the source page, author, license, and the attribution shown.
AI-generated illustrations
- When no suitable licensed photograph exists, we may use an AI-generated stylized illustration instead.
- Those illustrations are non-documentary: they are designed and checked so that they do not depict any identifiable real person, and are not presented as photographs of any real place or event. They are decorative and thematic, never evidence of anything.
- They are identified in two places: in the image’s alt text, which begins “AI-generated illustration:” so a reader using a screen reader is told before the description, and in a visible caption under the image reading “AI-generated illustration,” so anyone looking at the page can see what it is without inspecting the source.
- We do not use AI to generate an image of a real, identifiable person, and we do not use AI to depict a news event as though it were photographed.
What we never do with images
We do not alter the content of a news photograph, and we do not stage a photograph to look like documentation of an event. We take reasonable steps to verify the license of every image before use.
Independence and funding
Plainly Now is funded by the individual publisher who operates it. As of today the site carries no advertising, no sponsored posts, no affiliate links, no paid placements, and no paywall. Nobody pays us to cover a subject. No advertiser, sponsor, source, or subject of coverage has the right to approve, suppress, or alter our editorial conclusions. The site accepts no donations, grants, free products, travel, or payments from sources; the publisher has no financial interest in the companies or people covered; if any of that ever changes, it will be disclosed here.
If that ever changes — if the site takes advertising, accepts sponsorship, or introduces affiliate links — we will disclose it on this page and label the affected content, before it runs. Sponsored material, if it ever exists, will be labeled as sponsored and will not be written by the editorial process described above.
Contact
Corrections, questions about our method, and complaints all reach us at contact@plainlynow.com. Email is our only contact channel; we do not offer telephone support. The site is written for American readers, but it is operated from Latvia, in the European Union. See also our About page, our Privacy Policy, and our Terms of Service.
Last reviewed: July 30, 2026