Who writes Counsel.day, how, and what we will never do.
A public statement of editorial practice for everything on counsel.day. One author, James Graham, writes and signs every page; AI is used as a drafting and editing tool with every word read line by line before publication; sources are named where they exist and disclaimed where the author is interpreting rather than citing. This page also lists the things Counsel.day will never do, because the negative list is sometimes the most useful one.
One author, signed and accountable.
Every page on counsel.day is written by James Graham, the founder. There is no editorial team and no ghostwriting credit attached to any page. If a paragraph is wrong, that paragraph is the author's responsibility, and the contact for it is corrections@counsel.day.
The author is a data professional, not a clinician. The decision methodology in this site is grounded in his work in product analytics and engineering, not in clinical training; the framing is data-led, not therapeutic. Anything that reads as clinical claim is a writing error and should be reported as such.
AI as a drafting and editing tool, never as an accuracy substitute.
AI tools (currently Claude Opus 4.7) are used in two ways. First, as a drafting assistant: the author writes the brief and the structure; the AI produces a first-pass draft against that brief; the author rewrites the draft in his own voice and structure before the page ships. Second, as a copy editor: the final page is read back by the AI for inconsistencies, broken claims, and tone drift; the author resolves each flagged item one at a time.
What AI is not used for: fact verification, citation generation, original research, or claim substantiation. Every claim of fact on this site is checked against a named source by the author, by hand. If a source cannot be found or named, the claim is removed or rewritten as the author's opinion. AI hallucinations have shipped to other sites; they have not shipped to this one, because the AI is never the last reader.
Named sources, with the right disclaimers.
When the site refers to specific findings, those findings come from named, publicly available research. When it refers to clinicians or researchers by name (Daniel Kahneman on System 1 and System 2, John Gottman on dyadic disagreement, the broader couples-therapy literature on pseudo-agreement), the reference is to their published work, and to the methodology they inform; no clinician or researcher named on this site has endorsed Counsel.day, validated its method, or been involved in its construction.
That last sentence is the editorial line. Use of a named clinician's published work to inform our methodology is not endorsement; we will say so on the page where we reference them; we will correct any wording that suggests otherwise as soon as we are told about it.
Named corrections in the changelog.
If we publish something wrong, we fix it. Every correction is logged in the changelog with the date, the page, what was wrong, and what now reads. We do not silently update text. Substantive corrections (a factual claim, a price, a feature description, a named person's work) include a "corrected on" note on the page itself. Typographical corrections do not.
Write to corrections@counsel.day with the page URL and the specific text that is wrong. We acknowledge within one business day; we ship the correction within five business days for substantive errors, sooner if the error is misleading users.
The negative list.
Five hard commitments. These are not aspirational; they are constraints we have built the business around. If we break any of them, the user has grounds to walk; we expect to be held to all five.
- We will never sell or share your decision data. Not to advertisers, not to data brokers, not to insurance underwriters, not to "research partners". The data underneath a decision belongs to the people who made it.
- We will never use your decision data to train models. Not our own, not anyone else's. The Verdict AI runs against your data in a constrained execution environment that does not retain inputs or outputs beyond the verdict text itself, and that environment is firewalled from any training pipeline.
- We will never ad-fund the product. Counsel.day is paid for by users, in USD, on transparent subscription pricing. There are no ads, no sponsored content, and no affiliate deals embedded in the editorial.
- We will never claim therapeutic outcomes. The product is a data tool. It is not validated, built, tested, or endorsed by clinicians. We will not market it as a therapy substitute, a mental-health intervention, or an evidence-based treatment, because it is none of those things.
- We will never insert advertising into the verdict. The verdict view is the most important page on the site; it stays clean. No upsells, no cross-promotions, no banner asking you to invite friends. Just the answer the season produced.