AI Use and Writing Standards

published February, 2026

edited May, 2026

updated July, 2026

updated September, 2026

Why This Exists

When I publish under my name, issue professional evaluations or provide sworn testimony, I stand behind my words. They accumulate across time (I have left articles up on this site even if I no longer like the style, arguments or anger in them, as they are my intellectual record). My words carry reputational, professional and legal weight.

AI generates fluent language without bearing the responsibility of those words. These posted standards exists to prevent the abdication of intellectual, moral and professional responsibility while allowing legitimate structural and technical amplification.

I have found AI to be massively helpful in organizing my ideas, editing my work, conducting research and assisting with planning and analysis. Occasionally, the AI models I use (ChatGPT and Claude) try to suggest sentences and paragraphs. I reject them; they don’t sound like me. They offer up words that I don’t use in my writing. Even during the editing process, AI sometimes tries to smooth over my points. I do not accept it when it wants me to avoid controversy; however, I do accept that editorial advice when my writing puts me at legal risk.

I use AI extensively, but I do not use it the same way for every kind of work. I have allowed AI to do substantial drafting and rewriting on functional documents. By functional documents, I mean materials whose primary purpose is to accomplish an administrative, instructional or organizational task rather than to represent my writing or original thinking, such as: syllabi; CSWE competency mappings and accreditation materials; my student AI guide; and administrative clinical forms such as record releases, informed consent and emergency contact. In those cases, I provide the underlying facts, content, standards and professional judgment, review the work closely and remain responsible for signing off on the final product (for example, a LLM helped me with this paragraph).

My authored work is different. This includes articles published on this website or elsewhere, speeches, courses and other work in which the ideas, arguments and expression are presented as mine. I use AI for research assistance, fact-checking, hostile review, identifying weaknesses and contradictions and copy editing, but the ideas, experiences, arguments and writing must remain mine. Readers who spend their time reading an article published under my name, students taking one of my courses or audiences listening to me speak reasonably expect to encounter my thinking and my work. It is not enough for me to avoid impropriety; I must also guard against the appearance of impropriety.

Because of all that, I use AI tools under defined constraints.

This page explains those constraints.

Core Principles

1. Authorship Requires Ownership

The thoughts and words in work I present as my authored work are mine. I do not let AI brainstorm my arguments, positions or original ideas for me. To be clear, it can’t generate ideas like me. I do not let AI write my articles, speeches or other work that I present as my writing. Partly because it isn’t intellectually honest, and partly because it can’t write as well as me. Even as the technology improves, it can never be as good, because it can’t have physical experiences or understand love or grief or hope.

AI may draft functional documents, as defined above. It does not write my articles published on this website or elsewhere, my speeches or the substantive intellectual content of my courses.

AI does not generate my moral positions, clinical judgments, forensic opinions or final evaluative conclusions.

2. AI Has No Accountability

AI produces, at first glance, passable writing. But that writing is just a massive algorithm; for AI, there is no thought or meaning behind those words. Many people, from unknown to famous, believe they are seeing a ghost in the machine when they are chatting. They are not. AI writing does not carry authority, responsibility or understanding.

3. Moral and Forensic Claims Must Not Exceed Evidence

I have a standing order to the AI models for the following:

  • if I claim negligence exists, I may not imply intent
  • if evidence supports correlation, I may not assert causation
  • in forensic contexts, my conclusions must remain within documented findings and established standards; I may not make an emotional claim or explicitly urge a court on a particular path

This precision protects my credibility and helps me avoid potential legal issues. It also provides a check when I am angry. AI can help identify when my reaction to an event, institution or person is pushing my language or conclusions beyond what the evidence supports. In short, it is a failsafe to protect me reputatationally, professionally and legally from sudden responses based on perceived wrongs.

4. AI is a force multiplier

AI can strengthen expertise and efficiency, but it can also amplify shallow understanding and overconfidence. The value and reliability of my use of AI depend heavily on what I bring to the interaction.

If I use AI to discuss substance use treatment programs, I have decades of experience against which to evaluate what it tells me. I routinely identify bad assumptions, missing context and recommendations that sound reasonable but would not work in actual programs. The same is true, to varying degrees, when I use it to discuss Law Enforcement mental health, critical incident response, military and Veteran behavioral health, suicide or clinical supervision.

If I use AI to discuss something outside my expertise, such as lower back pain, I have far less ability to recognize a convincing error. I need more outside verification and should have less confidence in the interaction.

My confidence in AI should not be based simply on how knowledgeable or convincing it sounds. It should also depend on how qualified I am to recognize when it is wrong.

This is one of my concerns about students and workers using AI to perform tasks they have never learned to do themselves. AI may improve their immediate output while preventing them from developing new skills or eroding skills they already possess. It can make an experienced person more effective. It can also make an inexperienced person sound more knowledgeable than they actually are.

5. AI Should Challenge Me

I do not need AI to agree with me. Agreement is often the least useful thing it can provide.

My goal is to hone my use of AI so that it can serve, imperfectly, as some combination of George Orwell, my father, an in-house lawyer, hostile opposing counsel, an investigative journalist and a regular person who simply needs to understand my point.

Those roles serve different purposes. I want AI to identify unclear writing and weak thinking; hold me against my stated standards and goals; identify professional, reputational and legal risks; construct the strongest case against my position; find contradictions, missing evidence and inconvenient facts; and tell me when an ordinary reader would not understand or accept what I am saying.

I also use AI longitudinally. It can examine old articles, briefs, papers, reviews, notes, plans and previous discussions and compare them with what I say and do now. I can use that record to examine whether I am living up to my stated morals, work ethic, priorities and goals.

AI is not George Orwell, my father, a lawyer, opposing counsel, a journalist or an ordinary reader. It can misunderstand me, invent objections and give bad advice. Its criticism is something for me to evaluate, not a judgment I am required to accept.

The purpose is not to outsource my conscience or judgment. It is to make it harder for me to avoid an argument, fact, weakness or inconsistency simply because I would rather not confront it.

Permissible Uses

I may use AI tools for:

  • Organizing ideas I have already generated
  • Extracting themes, arguments and patterns from prior work, including articles, briefs, papers, reviews, notes and teaching materials
  • Comparing my current positions, decisions and behavior with my prior work, stated goals, morals, work ethic and standards
  • Identifying logical gaps, contradictions, overstatement, implied intent and unsupported causal claims
  • Mechanical editing, including spelling, grammar and repetition
  • Stress-testing clarity, structure and whether a nonspecialist can understand my point
  • Providing counterarguments and constructing strong challenges to my arguments, decisions and diagnostic reasoning
  • Assessing my writing and proposed actions for ethical, reputational, professional and legal risks
  • Identifying potentially relevant legal authorities, which I independently verify
  • Conducting research and identifying potentially useful journalism, academic research, government reports and other sources, which I verify as appropriate
  • Preparing background summaries on people, with consequential factual claims independently verified
  • Researching companies, organizations and institutions before meetings, engagements or professional decisions, with consequential factual claims independently verified
  • Assisting me in creating contemporaneous notes following significant meetings and interactions with people, businesses, clubs, organizations and institutions
  • Helping me evaluate my relationships with businesses and institutions, including compensation, workload, role, risk, opportunity cost and whether the relationship remains a meaningful use of my time
  • Helping me think through interactions with people outside my trusted professional circle, particularly new or peripheral professional contacts, prospective collaborators and people or organizations offering opportunities
  • Collating, comparing and formatting data
  • Refining clarity and efficiency in professional evaluations, with full de-identification
  • Preparing for court by organizing arguments, reviewing structure and stress-testing cross-examination vulnerabilities
  • Creating graphs and charts from government and other publicly available data for my classes and presentations
  • Maintaining and organizing bodies of articles, research and other source material for later use
  • Identifying conferences, speaking engagements and other professional opportunities
  • Researching prior conferences, speakers, audiences, organizations and other information relevant to deciding whether an opportunity is worth pursuing
  • Assisting with conference applications, speaker inquiries and other functional professional materials
  • Maintaining spreadsheets and other tracking systems for applications, projects and professional goals
  • Assisting with travel and hiking planning, including routes, mileage, elevation, water, transportation and contingency planning, subject to independent verification of safety-critical information
  • Identifying plays and other works I might want to read or see based on interests I have already developed
  • Serving as a sounding board for my reactions to plays, books and other works, including challenging my interpretations and comparing them with other works
  • Substantial drafting and rewriting of functional documents, including the specific categories identified above
  • Assisting with strategic planning, including career decisions, negotiations, pricing, workload, project priorities and opportunity costs
  • Assisting with course preparation through research, fact-checking, organization and review while leaving substantive teaching judgments and original course content to me
  • (a LLM had to point out a few of these uses to me when I presented this list)

Confidentiality, Clinical and Forensic Safeguards

  • No identifying client information is entered.
  • I do not enter confidential, privileged or otherwise protected professional information into AI systems unless the system and circumstances provide appropriate protections for that information.
  • No reconstruction of missing details occurs.
  • AI does not generate diagnostic conclusions, forensic opinions or sworn testimony.
  • AI may challenge my diagnostic or forensic reasoning, identify alternative interpretations and point out conclusions that exceed the available evidence.
  • Final opinions and testimony are mine alone.
  • All legal authorities are independently reviewed prior to use in court or publication.

AI may assist in editing for clarity and structural preparation. It does not participate in care or testimony.

Legal and Technical References

Where legal standards, statutes or case law are discussed:

  • AI tools may assist in identifying potentially relevant authorities.
  • All cited materials are independently reviewed prior to publication or testimony.
  • Responsibility for accuracy and interpretation remains solely mine.

Authorship and Intellectual Lineage

The historical, literary and cultural references in my writing and teaching reflect my own intellectual formation. I will not ask AI to generate historical, literary or cultural examples or references for me to insert into my writing, speeches or courses. I will not ask it to find a Twain reference that fits an argument I have already written or identify a television show popular with Gen Z so I can manufacture a connection with my students.

My parents and I have referenced Gatsby to each other my entire life. Orwell is my role model for political writing. Stan Lee helped shape my values. Shakespeare taught me to see complexity. It is of the utmost importance that I have lived and breathed the examples I use prior to writing, teaching or speaking about them.

AI may introduce me to books, plays and other works that I subsequently read or see. It may also serve as a sounding board when I discuss works I have actually experienced. Neither is a substitute for having engaged with the work myself.

Disclosure

When AI materially assists structural organization or legal reference identification in my authored or professional work, I will disclose that assistance.

I regularly use AI for research assistance, fact-checking, hostile review and mechanical editing of spelling, grammar and repetition. I do not routinely disclose those uses.

In my articles, speeches, courses and other authored intellectual work, the ideas, substantive references, arguments and writing are mine. As described above, AI may draft or rewrite functional documents. I review those materials and remain responsible for anything I submit, distribute or sign.

Scope

This framework governs my use of AI in:

  • Academic and professional writing
  • Speeches and presentations
  • Teaching and course preparation
  • Clinical skill refinement
  • Professional evaluations and forensic preparation
  • Research and fact-checking
  • Professional and strategic planning
  • Administrative and functional documents
  • Longitudinal review of my prior work, decisions, goals and standards
  • Travel, hiking and other planning
  • Intellectual exploration and discussion

It does not address broader societal or economic questions about AI. It establishes the professional standards I apply to my own work. AI has absolutely helped me organize and sharpen my work. I am very grateful that it did not appear when I was 15, because I would have used to do work for me and I would not have developed the same way.

Disclosure: Structural organization, editing, hostile reviewand risk-review assistance were provided by a large language model in accordance with my AI Use & Writing Standards. During revision, AI also identified several ways I have used it that I had not initially included. I reviewed, corrected and decided whether to include each one. All arguments, interpretations and conclusions are my own.
https://greenagel.com/ai-use-and-writing-standards/