How to write a psychological report: structure and limits

“Can you write me a report?” arrives from a school, a referring physician, an insurer and, every so often, from a lawyer preparing a custody case. Almost everything published about how to write one is a template — the easy part, and the least protective. What decides whether a report holds up is not its contents page but two prior questions: which document you are signing, and what you may assert inside it. This is documentation guidance, not legal advice.

A clinical report and a forensic opinion are not the same document

The APA’s Specialty Guidelines for Forensic Psychology (2013) define forensic practice in a way that surprises most clinicians: it does not turn on your subdiscipline or your job title. Their application “does not depend on the practitioner’s typical areas of practice or expertise, but rather, on the service provided in the case at hand.” You become a forensic practitioner through what you do in the case. They are aspirational by their own terms and “not intended to serve as a basis for disciplinary action or civil or criminal liability” — but they map the boundary better than anything else in print.

Guideline 4.02.02 draws the operative line:

“providing testimony on matters such as a patient’s reported history or other statements, mental status, diagnosis, progress, prognosis, and treatment would not ordinarily be considered forensic practice even when the testimony is related to a psycholegal issue before the decision maker. In contrast, rendering opinions and providing testimony about a person on psycholegal issues (e.g., criminal responsibility, legal causation, proximate cause, trial competence, testamentary capacity, the relative merits of parenting arrangements) would ordinarily be considered the practice of forensic psychology.”

Read the two halves against each other: you may describe what you observed and treated even when a court will use it, but not slip into an opinion on the legal question itself. Note where legal causation and proximate cause sit — on the forensic side. A clinician who writes “the accident caused the client’s depression” has crossed from describing a patient to deciding a case.

The pattern is not an American peculiarity: Mexico’s national code of criminal procedure reaches a similar result by statute. Article 369 requires experts to hold an official qualification in the field they opine on where that field is regulated, then exempts anyone testifying “as a witness about facts or circumstances they came to know spontaneously, even if in reporting on them they use the special aptitudes they possess in a science, art, technique or trade.” That is a criminal-procedure rule about who qualifies as an expert and who testifies as a witness, not a definition of the clinical role — a custody matter runs under a different code. Two systems, one recognisable boundary; verify how yours draws it.

If you are already their therapist, do not be their evaluator

Guideline 4.02.01 is explicit: providing forensic and therapeutic services to the same individual, or to closely related individuals, “involves multiple relationships that may impair objectivity and/or cause exploitation or other harm,” and practitioners asked or ordered to do both are encouraged to disclose the risk and refer elsewhere. Where referral is genuinely impossible, the guideline does not wave you through: it asks you to weigh the risks and benefits to all parties and to the legal system, to consider separating the two services widely in time, to seek judicial review and direction, and to consult knowledgeable colleagues. Standard 3.05(c) of the APA Ethics Code — which governs where law, institutional policy or extraordinary circumstances impose more than one role in a judicial or administrative proceeding — then requires clarifying role expectations and the extent of confidentiality at the outset and thereafter as changes occur. And Standard 2.01(f) adds a duty skipped constantly: on assuming forensic roles, psychologists “are or become reasonably familiar with the judicial or administrative rules governing their roles.”

Every report answers one question

The organising principle is the referral question, not the template. Standard 4.04(a) removes half the drafting problems in a line: reports include “only information germane to the purpose for which the communication is made,” and Guideline 11.04 says the same of padding. Get the question in writing before you accept. “A report on the client” is not a question. “Does the client show symptoms that interfere with occupational functioning, and what is the prognosis on current treatment?” is one, and it settles what stays out: the relationship history, the sibling’s diagnosis.

What must be true before you write

An examination adequate to what you intend to conclude. Standard 9.01(b) permits opinions on a person’s psychological characteristics “only after they have conducted an examination of the individuals adequate to support their statements or conclusions.” Where that is not practical despite reasonable efforts, you document the efforts and their result, “clarify the probable impact of their limited information on the reliability and validity of their opinions,” and limit the conclusions accordingly. This disposes of the most common failure in family reports: writing about the parent you never met. 9.01(c) is not the way around that: it is a narrow exception, for a records review where “an individual examination is not warranted or necessary for the opinion,” and it then requires you to say so and name your sources. Where an examination was warranted and you skipped it, a records review does not license the opinion.
Consent that covers the actual purpose. Standard 9.03(a) requires assessment consent to explain “the nature and purpose of the assessment, fees, involvement of third parties, and limits of confidentiality.” Standard 9.10 adds the counterpart: results are explained to the person unless the relationship precludes it — forensic evaluations among the examples — “and this fact has been clearly explained to the person being assessed in advance.” Where the law treats health information as sensitive data, as Mexico’s does, written consent is a statutory condition too — see informed consent in psychotherapy and, on the separate privacy notice that may be required, Mexico’s rules.
A written request before anything leaves your file. Record rules typically condition third-party disclosure on a written request from the client or their legal representative, and usually name a separate route for judicial and administrative authorities. Find the wording that binds you: a court order is not a client request, and the referral letter is rarely either.

The sections, read as reasoning

Few record rules prescribe the contents of a psychological report the way they prescribe the contents of a chart. Special-education evaluation reports are the usual exception, and where a list is prescribed it governs — check whether yours is one. Mexico shows the more common case: its clinical-record standard defines a resumen clínico whose minimum contents — presenting problem, diagnoses, treatments, course, prognosis and study results — make a serviceable skeleton, but the same numeral assigns it to a physician, and the norm permits in-house formats only on top of its own minimums. A format is a floor, never a defence. What follows is a reasoning checklist, not a legally sufficient template:

Identification of the requester, the person evaluated, and the signer.
The referral question, transcribed as it was put to you.
Sources and dates: sessions, instruments, documents, collateral informants.
Procedure: what you administered, and why that for this question.
Observations, attributed to their source.
Interpretation, separate from the observations.
Conclusions, bounded by the question.
Recommendations, if asked for.
Limitations.

Separating points 5 to 7 is not cosmetic. Guideline 11.02 asks practitioners to “distinguish observations, inferences, and conclusions,” and 11.03 to identify the source of each piece of information relied on — the habit behind a good SOAP note, and its absence among the costliest clinical documentation mistakes. Points 3 and 5 come from the intake interview and the chart; if the report accompanies the end of treatment, the reasoning is that of a documented termination.

A score is not a verdict

This is where reports overreach most often. Standard 9.02(b) conditions an instrument’s use on validity and reliability “established for use with members of the population tested,” and otherwise requires you to describe the strengths and limitations of the results. Standard 9.06 asks that interpretation account for situational, personal, linguistic and cultural differences and “indicate any significant limitations.” Standard 9.08(a) bars decisions grounded in results “that are outdated for the current purpose.” And 9.09(c) closes the technology excuse: psychologists “retain responsibility for the appropriate application, interpretation, and use of assessment instruments,” whether they scored the test themselves or used an automated service.

Translated into drafting: a screener reports a score and a band, not a diagnosis. The PHQ-9 and the GAD-7 screen and monitor; converting a cut point into a diagnostic category asserts more than the instrument supports. And a risk indicator calls for an immediate clinical risk assessment under your own protocol and training — the report documents that you did it, it does not stand in for it. The protocol comes from your professional guidelines and your training, not from an article.

A defensible findings paragraph reads like this:

“The PHQ-9 was administered on 14 March (16) and 9 May (9), by self-report, covering the fourteen days before each administration. Observation: on 9 May the client arrived on time, with no tearfulness, unlike the four preceding sessions. Inference: the drop is consistent with what was observed and with the client’s own report. Limitation: with no response-style measure and no collateral informants, the data come only from the client and my direct observation. This is a screening instrument, not a diagnosis.”

Limitations are a section, not a disclaimer

Those last two lines are the part most reports omit. Guideline 9.03 asks that where examining the person is not possible, practitioners make clear “the impact of such limitations on the reliability and validity of their professional products, opinions, or testimony.” A report with no limitations section reads not as stronger but as unexamined. Young sets the standard: “One overarching ethical principle that is especially applicable in forensic report writing is to be comprehensive, scientific, and impartial.”

The gap is measurable. A 2025 study in Acta Universitaria scored 60 Mexican forensic psychological reports and found a mean of 49.57 (SD = 12.84) across a range of 31.47 to 92.73, with the statement of hypotheses (M = 0.22) and instrument reliability and validity (M = 0.38) the weakest items. The evidence is bounded — child sexual abuse cases, 9 of 32 states, part of the sample recruited by convenience — not a portrait of clinical reports generally.

Formalities, and knowing when to decline

Whatever rule governs your records will want the same four things: the date, the full name of whoever wrote it, a signature, and retention for a defined minimum. Mexico’s standard, for example, requires date, time, full name and a handwritten, electronic or digital signature on every entry, and a minimum of five years from the last medical act. Your own jurisdiction’s minimum is a floor, not a target.

Declining is a clinical competence, not a lost sale. You decline when the question is psycholegal and your role is therapeutic; when asked to opine on someone you never examined; when the instrument that would carry the conclusion is not validated for that population; and when it needs forensic training you lack. Guideline 4.02.01 names the route: disclose the risk, and refer.

All of that depends on the chart already holding, dated and attributed, what the report will cite. gesell.ai keeps that base in order — intake, clinical history, treatment plan and session notes in a structured chart, with scales such as the PHQ-9 and GAD-7 scored automatically and dated — and turns your session record into notes you review and approve; it is also where the treatment plan with SMART goals behind most recommendations lives. The report and its signature stay yours: no tool decides for you what you are in a position to assert.