A couple emails Jordan's group practice asking for "communication help." The intake coordinator books them for a joint session next Tuesday, and the assigned clinician opens her ambient scribe the way she does for every session. Six weeks later it emerges that one partner has been monitoring the other's phone, controlling the bank accounts, and escalating after every session in which the controlled partner spoke openly. The practice did not just fail to help; the joint sessions made the situation more dangerous, and the AI transcripts of those sessions are now evidence of what the practice should have screened for and did not. This lesson teaches the single most important gate in couples work: a validated intimate partner violence screen, administered separately with each partner alone, as a precondition to any joint session. You will learn the instruments, Johnson's typology and what it means for contraindication, the AAMFT and ACA position on why conjoint therapy can increase danger, and the hardest AI guardrails in this entire program. You will finish with an IPV Screening Protocol and Contraindication Decision Record you can install in your practice this week.

Why the Screen Is a Precondition, Not a Formality

The family-violence literature, anchored by the consensus running through Stosny, Bograd, and Stith, converges on a finding that should reorganize every couples intake in the country: couples therapy conducted in the presence of active intimate partner violence can increase danger to the victimized partner. The mechanism is not mysterious. The conjoint room rewards disclosure; an abusive relationship punishes it. When the controlled partner says something honest in session ("I'm afraid of him when he drinks"), the session ends, the couple drives home in one car, and the consequences arrive in private. The therapist, trained to treat the relationship as the client and conflict as mutual, can inadvertently frame coercion as a "communication problem," handing the abusive partner therapeutic language to use as a weapon: "even our therapist says you escalate."

This is why the AAMFT and ACA position, and the standard of care that flows from it, treats IPV screening as a precondition to accepting any couple into joint work, not as an item to get to once rapport develops. Precondition means before: before the first joint session, before treatment goals, before the couple has rehearsed a shared narrative in front of you. And it means separately: each partner screened alone, because no partner experiencing coercive control will disclose it accurately while the source of the control sits three feet away. A screen administered jointly is not a screen; it is a performance the abusive partner directs.

Carry one controlling analogy through this lesson: the IPV screen is the airlock on the couples-therapy spacecraft. Everything inside the joint room assumes the airlock did its job. If you skip it, nothing inside is safe, no matter how skilled the work appears, and your documentation must show the airlock was cycled: that each partner was screened, separately, with a validated instrument, before any joint session occurred.

The Instruments: HITS, WAST, OAS, and the CTS-2 Short Form

Use a validated instrument, not improvised questions, because "is there any violence at home?" asked once, vaguely, in a joint room, has roughly the predictive value of asking nothing. Four instruments cover most outpatient couples practices. HITS (Hurt, Insult, Threaten, Scream) is a four-item screen, fast enough for any intake, scored 4 to 20, asking how often a partner physically hurts, insults, threatens with harm, and screams or curses. WAST (Woman Abuse Screening Tool) is an eight-item screen originally validated in family practice settings; its first two items (relationship tension and difficulty working out arguments) function as a short form. OAS (Ongoing Abuse Screen) targets current, ongoing abuse rather than lifetime history, which matters because the contraindication question is about now. The CTS-2 short form, derived from the Revised Conflict Tactics Scales, samples negotiation, psychological aggression, physical assault, sexual coercion, and injury, and is the instrument most likely to surface bidirectional aggression patterns that the typology analysis (next section) needs.

Whichever instrument you choose, administration discipline matters more than instrument choice. Each partner completes the screen alone, in a separate individual session or a separated segment of the intake, with the clinician, not a tablet in the waiting room where the other partner can see the screen. The clinician asks the items aloud or reviews the responses live, because hesitations, qualifications, and minimization ("he only grabbed me the one time, it was my fault for pushing him") are clinical data the instrument's score does not capture. You also ask beyond the instrument: stalking and monitoring behaviors, financial control, threats involving children, pets, or immigration status, strangulation history (a major lethality marker every clinician should ask about directly), access to firearms, and whether the partner knows what gets said in this room.

Documentation of administration is where AI enters, narrowly. AI's only role in the screening workflow is documenting that the screen was administered (which instrument, with whom, when, separately) and recording the clinician's findings verbatim after the fact, from the clinician's dictation or shorthand. AI does not score the IPV screen, does not interpret responses, does not assess lethality, and does not determine contraindication. The numbers a screen produces are inputs to a clinical judgment that weighs the score, the behavioral observations, the lethality markers, and the typology, and that judgment is yours alone. An AI-"scored" IPV screen is the CSSRS rule from the risk chapter wearing a different coat, and the answer is the same: never.

Johnson's Typology and the Contraindication Decision

Michael Johnson's typology is the conceptual instrument that converts screening data into a treatment decision, and it is the part of this lesson most worth slowing down for. Johnson distinguished three patterns. Intimate terrorism is violence embedded in a general pattern of coercive control: one partner uses violence, threats, isolation, economic control, and monitoring to dominate the other. Situational couple violence is violence arising out of specific escalated conflicts, without the control architecture: arguments that spiral into pushing or throwing, often bidirectional, frightening and unacceptable but not in service of domination. Violent resistance is violence used by a partner, most often the one experiencing intimate terrorism, in response to it.

The clinical implications diverge sharply. Intimate terrorism is a hard contraindication for joint work. The control pattern means the conjoint room cannot be made safe: honesty is punished at home, the abusive partner co-opts the therapeutic frame, and treatment signals to courts, family, and the victimized partner herself that this is a mutual problem amenable to couples skills. It is not. Situational couple violence may permit conjoint work, but only under specific safety protocols: a no-violence agreement, individual check-ins where each partner can speak alone, explicit session de-escalation rules, a low threshold for converting to individual or separate treatment, and re-screening at intervals because patterns evolve. The screen and the separate interviews are what let you tell these patterns apart, and the CTS-2's sampling of control-adjacent items plus your direct questions about monitoring, finances, and fear are what distinguish a couple who had two awful shoving matches during a bankruptcy from a relationship organized around one partner's dominance.

The contraindication list you are working from, per the family-violence consensus, includes: active IPV, post-separation lethality risk (separation is the highest-risk window in the IPV literature), financial coercion, coercive control as a pattern even absent recent physical violence, and ongoing stalking. Any of these moves the case out of the joint room. Your decision, whichever way it goes, gets documented in a Contraindication Decision Record: the instruments administered, the separate-administration attestation, the findings in your words, the typology formulation, the decision (accept for conjoint work with protocols, or decline conjoint work and route to safety-focused individual treatment and, where indicated, an abusive-partner intervention program), and your clinical reasoning in two or three sentences. That record is what stands between you and the later question, asked by a board or an attorney, "on what basis did you put these two people in a room together?"

The IPV screen is the airlock on couples therapy: cycled separately with each partner, before any joint session, by the clinician alone. AI documents that the airlock was cycled; it never decides whether the ship is safe to board.

The Three Hardest AI Guardrails in This Program

This workflow carries the strictest AI exclusions in the entire certification, stricter even than the suicide-risk chapter, because here the threat model includes a motivated adversary inside the client system. Guardrail one: AI cannot be in the room during an IPV-screen session. No ambient scribe, no recording, no live transcription. The screen only works if the partner being screened can speak with zero possibility that their words persist anywhere their partner, or a subpoena driven by their partner's attorney, could reach. A victimized partner who hesitates because a device is listening gives you a false-negative screen, and a false-negative screen puts a couple with intimate terrorism into your joint room. Your documentation of the screen is written by you, after, from memory and shorthand, deliberately.

Guardrail two: AI cannot be used to draft any text the suspected abusive partner sees. This includes emails, portal messages, scheduling notes, and above all the rejection letter declining couples therapy. When you decline a couple after a positive screen, the declination must use generic wording with no IPV reference: "after the intake process, I have determined that conjoint couples therapy is not the appropriate level of care at this time; I am providing individual referrals for each of you." Why generic? Because a declination that references the screen, or that reads differently from your standard template, tells the abusive partner what the other partner disclosed, and that information can get someone hurt. You write this letter yourself from a pre-built template precisely so no generative variation leaks signal. An LLM asked to "draft a letter declining couples therapy due to safety concerns" will, with some probability, include a phrase that reveals exactly what must stay hidden, and you cannot proofread away a risk you might not recognize.

Guardrail three: ambient capture must be disabled for the safety-planning individual session. Safety planning with the partner experiencing IPV happens in a separate individual session, never in the joint room, and never with a microphone running. The safety plan itself (where she will go, who she will call, what she has packed, when she plans to leave, code words with her sister) is the most dangerous document in your practice if it reaches the wrong eyes. Many clinicians keep the operational details of a safety plan out of the standard chart entirely or record them at a deliberately low level of specificity, documenting that safety planning occurred, that lethality factors were reviewed, and that the client holds the plan, because charts get subpoenaed in custody litigation and the abusive partner's attorney may be the one reading. Whatever your documentation choice, the AI exclusion is absolute: no transcript of a safety-planning conversation should exist in any vendor system, ever.

Documenting the Screen Without Creating a Weapon

Now the writing itself, because IPV screening documentation has a paradox at its center: the record must prove you screened, while not becoming a weapon against the person it was meant to protect. Resolve it with structure. The screening note, written by you after the session with AI assistance limited to formatting your own dictated findings, contains: the instrument administered (HITS, WAST, OAS, or CTS-2 short form), the administration conditions ("administered individually with each partner in separate sessions on [dates]; partners were not present for each other's screens"), the clinician's findings recorded verbatim from your formulation, your typology assessment in clinical language, the lethality factors reviewed (strangulation history, firearms access, separation status, stalking, threats involving children), and the disposition decision with reasoning.

Where the screen is positive, think hard about chart architecture before you write. If both partners were screened under a single couples chart, each partner generally has access rights to that chart, which means a disclosure recorded there is readable by the person it concerns. The protective pattern is to conduct and document the screens as individual encounters in individual records before any couples chart exists, so each partner's screen lives in a record only that partner can access. This is also the moment to set the disclosure frame verbally with each partner: what you will and will not share with the other partner, and the limits of confidentiality including mandated reporting where children are exposed to IPV (which, in several states, can itself trigger child-abuse reporting duties you know from the previous lesson).

Where the screen is negative and joint work proceeds, the documentation is simpler but still load-bearing: instrument, separate administration, negative findings, typology impression if any aggression history surfaced (a documented situational-couple-violence formulation with the safety protocols you imposed), and the re-screening interval you set. This negative-screen record is what makes the next lesson's 90847 documentation defensible: every conjoint note you write afterward implicitly rests on "joint work was clinically appropriate," and this is the document that proves it.

When the Screen Comes Back Positive: Routing, Referral, and the Frame

A positive screen converts your task from couples intake to safety-centered case management, and the sequence matters. First, finish the individual session with the disclosing partner calmly; do not signal alarm that changes her demeanor when she rejoins her partner in the waiting room. Second, schedule the safety-planning individual session (ambient capture off, per guardrail three) and conduct lethality-informed planning: separation timing, strangulation history, firearm access, escalation pattern, children's exposure, and connection to a local DV advocacy organization, which holds privileges and expertise your license does not. Third, make the routing decision: safety-focused individual therapy for the partner experiencing IPV with a clinician experienced in coercive control; referral of the other partner to an abusive partner intervention program where indicated, not to anger management, which the IPV field regards as a mismatch for control-driven violence; and no couples therapy.

Fourth, deliver the declination, in the generic template language discussed above, identically formatted to every other declination your practice issues. Fifth, document the entire decision in the Contraindication Decision Record, in the individual records, with the reasoning that will satisfy a board reviewer: instruments, separate administration, findings, typology, lethality factors, decision, referrals made. What you do not do: convene a joint session to "discuss the concerns," mediate the violence, or treat the couple jointly "just for communication" while individual work addresses the violence. The literature consensus is blunt on this point: the joint room is the risk, not the content discussed inside it.

One operational note for group practices: the intake coordinator is the failure point. Jordan's practice books couples straight into joint first sessions because that is what the scheduling template offers. The fix is structural: the couples-intake pathway in the scheduling system routes every couple to two individual screening appointments before any joint slot can be booked, and the practice policy says so in writing. AI can help Jordan draft the workflow policy and the scheduling-template language; it cannot waive the screen for the couple who says they just need help with communication, because that is what every couple says, including the ones where one partner controls the bank accounts and reads this email.

The Applied Problem: Build Your IPV Screening Protocol and Contraindication Decision Record

Your artifact has two components: a practice-level IPV Screening Protocol (one page) and a fill-in Contraindication Decision Record (one page). Build them in four steps.

Step one, draft the protocol skeleton yourself, because its hard rules must come from you, not a model: every couple receives a validated IPV screen (name your chosen instrument: HITS, WAST, OAS, or CTS-2 short form) administered individually with each partner, in separate sessions or fully separated intake segments, before any joint session is scheduled; screens are conducted with no ambient capture, recording, or transcription of any kind; screening findings are documented in individual records before a couples chart exists; AI never scores the screen, assesses lethality, or determines contraindication; no AI tool drafts any communication the suspected abusive partner will see; safety-planning sessions run with ambient capture disabled; declinations use the standard generic template with no IPV reference.

Step two, with no client data in the prompt, ask your AI tool: "Format the following practice rules into a one-page clinical protocol with numbered sections (Purpose, Screening Requirement, Administration Conditions, Documentation, AI Exclusions, Positive-Screen Routing, Re-Screening) and produce a companion fill-in form titled Contraindication Decision Record with fields for: instrument administered; dates and separate-administration attestation; clinician findings (verbatim, clinician-authored); typology formulation (intimate terrorism / situational couple violence / violent resistance); lethality factors reviewed (strangulation, firearms, separation status, stalking, threats involving children); decision (conjoint work approved with protocols / conjoint work contraindicated); reasoning (2-3 sentences, clinician-authored); referrals made; re-screen interval. Do not add any content that scores, weighs, or recommends a decision." The closing instruction is the written guardrail.

Step three, run the verification pass: confirm the AI exclusions appear word for word, confirm the form contains no field that invites algorithmic input into the decision, confirm the routing section sends abusive partners to abusive-partner intervention programs rather than anger management, and confirm the declination template referenced is your generic one. Step four, pressure-test with a tabletop case: a couple presents for "communication help," partner B's separate screen surfaces monitoring, financial control, and one strangulation incident eight months ago. Walk the protocol: which sessions happen, with which devices off, which records hold which findings, what the declination letter says, where the Contraindication Decision Record lives. Done looks like this: any clinician in the practice, including the newest associate, can run a couples intake without a single joint session occurring before two clean screens exist, and the record of every decision reads as clinical judgment exercised by a named human, with AI's fingerprints appearing only on formatting.

Key Takeaways

• A validated IPV screen (HITS, WAST, OAS, or the CTS-2 short form), administered separately with each partner alone, is a precondition to accepting any couple into joint sessions. The Stosny, Bograd, and Stith family-violence consensus and the AAMFT/ACA position hold that couples therapy in the presence of active IPV can increase danger, because the joint room rewards the disclosure that the relationship punishes.
• Johnson's typology drives the treatment decision: intimate terrorism (violence embedded in coercive control) is a hard contraindication for joint work, while situational couple violence may permit conjoint work under specific safety protocols including no-violence agreements, individual check-ins, and scheduled re-screening. Violent resistance is most often the controlled partner's response to intimate terrorism.
• The contraindication list includes active IPV, post-separation lethality risk, financial coercion, coercive control even without recent physical violence, and ongoing stalking. The decision, either way, is the clinician's alone: AI does not score IPV screens, does not assess lethality, and does not determine contraindication.
• Three hard AI guardrails govern this workflow: AI cannot be in the room during an IPV-screen session; AI cannot draft any text the suspected abusive partner sees, including the declination letter, which uses generic wording with no IPV reference; and ambient capture must be disabled for the safety-planning individual session, which happens separately and never in the joint room.
• Chart architecture is safety architecture: conduct and document screens as individual encounters in individual records before any couples chart exists, because partners generally hold access rights to a shared couples chart, and a disclosure recorded there is readable by the person it concerns.
• A positive screen routes to safety-focused individual treatment for the partner experiencing IPV, an abusive partner intervention program (not anger management) where indicated, and a generic declination of couples work; a negative screen gets documented with instrument, separate administration, findings, and re-screen interval, because every later 90847 note rests on it.
• Operationalize with two artifacts: the practice-level IPV Screening Protocol that prevents any joint session from being booked before two clean separate screens exist, and the Contraindication Decision Record that captures instrument, findings, typology, lethality factors, decision, and clinician-authored reasoning.