An I-130 intake used to eat 90 minutes. A well-scoped prompt now gets you a first-draft summary in eight. That is the entire promise of ChatGPT for immigration practice in 2026, and also the entire risk. The output looks confident. It cites regulations that sometimes do not exist. It happily fills gaps a paralegal would flag. This guide gives you 15 prompts that survived real caseload use, plus the compliance guardrails that keep you on the right side of ABA Rule 1.1 and every state bar unauthorized-practice statute you answer to.
What immigration lawyers can actually automate with ChatGPT
The honest answer is narrow. ChatGPT prompts for immigration lawyers in 2026 work well on structured, repeatable drafting where you already know the correct answer and just need faster typing. They fail on anything requiring current USCIS policy memos, priority date math against the latest Visa Bulletin, or country-specific consular quirks that changed last Tuesday.
What actually saves time in a busy practice:
- First-pass intake summaries from a 30-minute consult recording or notes
- Plain-English explanations of visa timelines for clients who Google everything at 2 a.m.
- Scaffolds for RFE responses (structure, not substance)
- Consular processing checklists tailored per post
- Cover letters for adjustment-of-status packets
- Motion-to-reopen skeletons pointing at the right regulatory hooks
- Interpreter and translator briefing notes
- PERM prevailing-wage narrative drafts
- DACA renewal reminder sequences with the 150-day window baked in
- Citizenship interview prep questionnaires by civics score band
What does not work: legal research on current policy, filing-fee lookups, priority date predictions, or anything a client will rely on without your review. Treat every output as a paralegal draft from someone on their first day.
Build a private prompt library your firm owns. Store finalized prompts in a shared workspace, version them by USCIS form edition, and log which associate approved each template. When a form revises, you update once. See the PromptSpace library for organizing patterns.
15 copy-paste prompts organized by task
Every prompt below uses bracketed slots. Fill them before running. Never paste client identifiers, A-numbers, passport numbers, or dates of birth into a consumer ChatGPT account — see the compliance section for why that alone can end a career.
1. Family-based intake summary (I-130 / I-485)
You are drafting an internal intake memo for an immigration attorney. Below are notes from a 30-minute consult with a prospective client seeking a family-based petition. Produce a memo with these sections: (1) Parties and Relationship, (2) Petitioner status and eligibility, (3) Beneficiary immigration history and any inadmissibility flags, (4) Priority date category estimate, (5) Consular vs adjustment recommendation with reasons, (6) Open questions the attorney must answer before quoting fees. Flag any inconsistency in dates. Do not invent facts. Where a field is missing write MISSING - NEEDS FOLLOW-UP.
Consult notes: [PASTE ANONYMIZED CONSULT NOTES HERE]
2. Employment-based intake (EB-2 / EB-3 / EB-1)
Act as senior paralegal for a US business-immigration firm. From the notes below, draft an internal case-strategy memo covering: (1) Beneficiary credentials and equivalency posture, (2) Sponsoring employer profile and ability-to-pay signals, (3) Likely category (EB-1A, EB-1B, EB-2 NIW, EB-2 PERM, EB-3 skilled), (4) PERM feasibility and prevailing-wage risk, (5) Priority date realism given the beneficiary's country of chargeability, (6) Timeline estimate with key gates. Keep every recommendation contingent on attorney review. Note any facts that would trigger a public-access-file requirement.
Notes: [PASTE ANONYMIZED NOTES]
3. Asylum intake and declaration outline
You are helping an immigration attorney prepare an internal outline for an affirmative asylum case (Form I-589). From the intake notes below produce: (1) Protected ground analysis (race, religion, nationality, political opinion, PSG) with a two-sentence theory of the case for each ground that might apply, (2) One-year filing deadline analysis and any changed or extraordinary circumstances, (3) Corroborating evidence checklist tailored to the country of feared persecution, (4) Country-conditions research questions the attorney should assign, (5) Draft outline for the client declaration in first person, using ONLY facts stated in the notes. Do not embellish. Mark every inferred detail INFERRED - CONFIRM.
Intake notes: [PASTE ANONYMIZED NOTES]
4. Plain-English visa timeline explainer
Rewrite the timeline below as a calm, plain-English email to a client who is anxious and checks USCIS case status three times a day. Use short sentences. No legal jargon. Explain what each step means, what could delay it, what the client can and cannot do to speed it up, and what silence from USCIS actually means at each stage. End with a single sentence about when they should contact our office. Do not promise dates. Reading level: eighth grade.
Case type: [E.G. I-130 SPOUSE OF US CITIZEN, CONSULAR PROCESSING] Country of chargeability: [COUNTRY] Filing date: [DATE] Current stage: [STAGE]
5. RFE response scaffold (attorney review required)
You are drafting a STRUCTURAL SCAFFOLD only for an attorney response to a USCIS Request for Evidence. Do not write substantive legal argument. Produce: (1) A section-by-section outline mirroring each RFE issue, (2) For each issue, a bullet list of the exact document types typically responsive, (3) A placeholder legal-argument paragraph marked ATTORNEY TO DRAFT with the elements the attorney must address, (4) An evidence index template, (5) A cover-letter skeleton addressed to the correct service center. Insert NO regulatory citations, statute numbers, or case names — the attorney will add these after verifying current authority.
RFE issues (paraphrased, no client PII): [LIST ISSUES] Form type: [E.G. I-140 EB-2] Service center: [SERVICE CENTER]
6. Consular processing checklist by country
Produce a consular processing checklist for an immigrant visa interview at the US Embassy or Consulate in [COUNTRY / POST]. Structure: (1) Pre-interview NVC document upload list, (2) Civil documents required from this specific country (marriage, birth, police, military) with known naming or authentication quirks, (3) DS-260 fields that commonly trigger 221(g) at this post, (4) Medical exam panel physician logistics, (5) Interview-day document originals, (6) Post-interview steps including visa issuance timing and port-of-entry reminders. Flag every item where post-specific practice is known to vary and mark VERIFY WITH POST. Do not invent embassy procedures.
Case type: [IR-1 / CR-1 / F2A / EB-3 etc.]
7. Adjustment-of-status packet cover letter
Draft a professional cover letter to USCIS for a concurrent I-130 and I-485 filing (adjustment of status based on marriage to a US citizen). Include: firm letterhead placeholder, correct lockbox address placeholder [VERIFY CURRENT ADDRESS], list of forms enclosed, list of supporting evidence organized by tab, filing fees table with placeholders, and a courteous request for receipt notices. Keep the tone factual and short. Do not summarize the merits of the case. End with attorney signature block. Mark every dollar amount and address as [VERIFY CURRENT].
Case-specific facts to reference (no PII): [LIST NON-IDENTIFYING FACTS]
8. Motion to reopen template (structural)
Produce a structural template for a Motion to Reopen before [USCIS / EOIR - SPECIFY]. Sections required: (1) Caption and case identifiers, (2) Statement of jurisdiction and timeliness with placeholders for filing-deadline math, (3) Procedural history skeleton, (4) Statement of new facts placeholder marked ATTORNEY TO DRAFT, (5) Argument framework with headings only, (6) Prayer for relief, (7) Certificate of service. Include NO citations, NO regulatory subsections, and NO factual assertions. This is scaffolding for attorney completion. Note any deadline the attorney must independently verify against the current regulations.
9. Interpreter and translator briefing
Draft a pre-hearing briefing note for a court interpreter working on a [LANGUAGE] asylum merits hearing. Cover: (1) The interpreter's role and neutrality obligations, (2) Sensitive subject-matter warning without disclosing case facts, (3) Terminology the interpreter should verify in advance (asylum, well-founded fear, particular social group, credible fear, etc.), (4) Protocol for interruptions and clarifications, (5) Confidentiality reminder and document handling. Keep it under 400 words. Add a second briefing note for a document translator handling civil documents with a certification-of-translation template at the end.
10. PERM prevailing-wage narrative draft
You are drafting an internal narrative to support a PERM prevailing-wage request (Form ETA-9141) and, later, the ETA-9089. From the position facts below produce: (1) Position summary in duties-first order matching O*NET verbiage, (2) Minimum requirements narrative expressed as objective, business-necessity-defensible criteria, (3) Alternate requirements discussion with Kellogg language considerations, (4) Special skills justification tied to actual duties, (5) A red-flag list of any requirement that appears tailored to the beneficiary and would create a business-necessity problem at audit. Do not assert prevailing-wage levels — that is DOL's determination.
Position facts: [JOB TITLE, DUTIES, EDUCATION, EXPERIENCE, WORKSITE, TRAVEL, SUPERVISION]
11. DACA renewal reminder sequence
Draft a four-email reminder sequence for a DACA renewal client. Timing: 180 days before EAD expiration, 150 days (the recommended filing window opens), 120 days, and 90 days. Each email must: (1) State days remaining in plain language, (2) List documents the client should gather, (3) Explain what happens if they miss the window, (4) Include a one-click reply option to schedule filing, (5) Avoid any language that could be read as legal advice to non-clients. Tone: warm, direct, no alarmism. Add a fifth email template for the case where the client goes silent past the 90-day mark.
12. Citizenship interview prep questionnaire
Create a citizenship interview preparation questionnaire for an N-400 applicant. Sections: (1) 20 personal-history questions the officer commonly asks in plain English, mapped to the exact N-400 part and question number, (2) A civics practice set of 20 questions drawn from the current USCIS 100-question pool — mark the set VERIFY AGAINST CURRENT USCIS LIST before use, (3) A good-moral-character review checklist covering the statutory period, (4) Red-flag topics that require attorney consultation before the interview (tax filings, selective service, arrests, prior immigration violations), (5) Day-of logistics reminder. Reading level: seventh grade.
13. Post-consult follow-up email
Draft a follow-up email to send within 24 hours of an initial consult. Include: (1) A one-paragraph recap of the case type discussed without restating client PII, (2) A clear next-step list with owners (client vs firm), (3) Fee-agreement transmission note if applicable, (4) A short section on documents the client should begin gathering, (5) Contact protocol expectations. Warm but not chatty. End with a sentence confirming that no attorney-client relationship exists until a signed engagement letter is on file.
Case type discussed: [TYPE] Next step decided: [STEP]
14. Country-conditions research question set
Generate a country-conditions research question set for an asylum case involving [COUNTRY] and the protected ground of [GROUND]. Produce: (1) 15 targeted research questions a paralegal should answer using State Department reports, UNHCR, HRW, Amnesty International, and reputable news, (2) A source-quality rubric, (3) A template for a country-conditions memo tying findings back to the client's specific fear, (4) A list of expert-witness categories that might be needed. Do not invent country facts. The paralegal will source everything.
15. Fee-agreement plain-English addendum
Write a plain-English one-page addendum that sits alongside the formal engagement letter. Cover in short paragraphs: (1) What our firm will do, (2) What we will NOT do (no guarantees, no policy changes, no case-status miracles), (3) How fees are structured for this case type, (4) When additional fees apply, (5) How the client should communicate with us, (6) What happens if the client's case is denied. Reading level: eighth grade. Do not replace or contradict the formal engagement letter — this is a comprehension aid only.
For every prompt above, run it once with fake facts, review the output structure, then save your firm's approved version. Prompts drift. Model updates change tone. Your library needs an owner and a quarterly review cadence, just like your form templates.
ChatGPT vs Claude vs Gemini for immigration practice
Different tools, different strengths. The table below reflects real workflow patterns across US immigration firms in 2026. Pick based on your primary bottleneck, not on brand loyalty.
| Task | ChatGPT (GPT-5 tier) | Claude (Sonnet / Opus) | Gemini (2.x) |
|---|---|---|---|
| Intake memo from notes | Strong, concise | Strongest at structure and nuance | Good, faster on very long inputs |
| Long-document review (RFE, NOID) | Solid to 128k | Best for 200k+ token records | Best for 1M+ token dumps |
| Plain-English client emails | Warmest default tone | More formal, easier to tune down | Serviceable, needs style prompt |
| Regulatory citations | Hallucinates — do not trust | Hallucinates less, still verify | Hallucinates — do not trust |
| Country-conditions synthesis | Good with web search on | Good with web search on | Strongest live-web integration |
| PII risk on consumer tier | High — training on by default on free plan | Lower — no training on API | High on consumer tier |
| Enterprise / HIPAA-adjacent tier | ChatGPT Enterprise / Team | Claude for Work / API zero-retention | Gemini Enterprise via Workspace |
The practical rule: use enterprise tiers with zero-retention terms for any prompt that comes within a mile of a real client. Free consumer accounts are for practicing prompt syntax on made-up cases, nothing else. See OpenAI's enterprise privacy terms and Anthropic's commercial terms before signing anything.
USCIS itself uses machine-assisted review in some adjudication pipelines. That does not lower your duty of competence on the other side of the counter. It raises it. If the agency's tools flag a pattern in your filings, you want your own tools to have caught it first.
Honest limitations: what ChatGPT cannot do for immigration lawyers
This section is the reason this article exists at 15 minutes instead of five. Immigration is the highest-stakes vertical in consumer legal AI use because a wrong output can separate a family, lose a career-defining case, or expose the attorney to bar discipline. The failures below are not hypothetical.
Hallucinated regulatory citations
ChatGPT will cheerfully cite 8 CFR sections that do not exist, invent BIA decisions, and misattribute holdings. In 2023 a New York attorney was sanctioned in a civil case for filing ChatGPT-generated fake citations. Immigration practitioners face the same risk in front of USCIS, the EOIR, and federal courts. Every citation the model produces requires independent verification against the current CFR, USCIS Policy Manual, and case databases. There is no shortcut here.
Stale training data
Visa Bulletin movements, filing addresses, fee schedules, policy memos, and executive orders all change on schedules the model does not track. A prompt asking for the current I-485 filing fee will confidently give you a number that was correct two years ago. Never let a client-facing number pass through without a fresh USCIS.gov check.
Country-specific consular quirks
Post-by-post practice varies. The consulate in Mumbai handles CR-1 interviews differently than the one in Chennai. The model does not know last month's operational change at your specific post. Treat any country-specific output as a starting checklist, then verify with the local attorney bar or a colleague who filed there recently.
No judgment on discretionary factors
Discretion runs through immigration practice — hardship analysis, good moral character, humanitarian factors, prosecutorial discretion. The model produces plausible paragraphs on these topics. It cannot weigh the actual facts of your client's life. That is your job and your license on the line.
Compliance and ethics — read every line.
ABA Model Rule 1.1 (Competence): Comment 8 requires lawyers to keep abreast of the benefits and risks associated with relevant technology. In 2026 that explicitly includes generative AI. If you use ChatGPT in your practice without understanding its failure modes, you are arguably out of compliance with your baseline duty of competence. Read your state's version of Rule 1.1 and any 2024-2026 formal ethics opinions on AI (California, Florida, New York, and the ABA itself have all issued guidance). See ABA Rule 1.1.
Unauthorized practice of law (UPL): ChatGPT is not licensed. If you give a client an AI-generated answer without meaningful attorney review, you are effectively letting an unlicensed entity practice law on your letterhead. State UPL statutes reach further than most attorneys realize. Non-attorney staff using these prompts must operate under attorney supervision with real review, not rubber-stamp sign-off.
Client confidentiality (Rule 1.6): NEVER paste A-numbers, passport numbers, dates of birth, addresses, employer names, or any client-identifying information into a consumer ChatGPT account. Free-tier prompts may be used for model training. Even on paid consumer tiers, retention terms differ from enterprise. Use ChatGPT Enterprise, ChatGPT Team, or an API integration with zero-retention terms — and get the DPA signed before the first real prompt. If your firm handles cases touching federal contracts or classified backgrounds, add a second layer of counsel on data handling.
Mandatory attorney review: Every output from every prompt in this article requires an admitted attorney to read it end to end before it leaves the firm, before it goes into a filing, and before it reaches a client. The prompts are drafting accelerators, not decision-makers. Treat them exactly like a first-year associate's first draft — useful, dangerous, and never final.
Language and cultural nuance
The model's translations and cultural framings drift. In asylum work especially, the difference between the client's actual words and a smoothed-over English rendering can be the difference between a grant and a denial on credibility. Use certified interpreters and translators. Use the model for briefing them, not for replacing them.
Workflow integration: embedding prompts in your practice
A prompt library only pays back if the firm uses it consistently. The pattern that works across small-to-mid immigration firms in 2026 looks like this.
Weekly rhythm
- Monday morning: paralegal runs intake summaries for the week's new consults using the approved template
- Attorney reviews summaries before consult follow-ups go out (mandatory sign-off)
- Mid-week: RFE and NOID responses drafted with the scaffold prompt, then handed to the attorney with a highlighted citation checklist
- Friday: DACA and renewal reminder sequences reviewed and sent through the CRM
- End of month: prompt library owner runs a sample audit — 10 random outputs, checked for accuracy drift
Tooling stack that keeps you compliant
- ChatGPT Enterprise or Team account (never free) with SSO and audit logs
- Prompt library in a shared workspace with version control
- Case-management system (Docketwise, INSZoom, MyCase) as the system of record — the AI outputs never live there without attorney sign-off
- A DPA and BAA-equivalent on file with every AI vendor
- An internal one-page AI policy every associate and paralegal signs annually
Add a required checkbox in your case-management workflow: "AI-assisted draft — attorney reviewed [initials] [date]." It creates an audit trail. It also forces the human moment that prevents most of the horror stories.
Training your team
The prompt library is only half the work. The other half is training everyone who touches it. Run a 60-minute onboarding covering: what these prompts are, what they are not, what data can never enter them, and what attorney review means in practical terms. Repeat every six months. Track completion. When the next state bar ethics opinion drops, update the training the same week.
For adjacent professional-services playbooks, see ChatGPT prompts for financial advisors in 2026 and best AI tools for CPA firms in 2026. The compliance patterns transfer. The domain risk does not.
FAQ: ChatGPT prompts for immigration lawyers
Is it ethical to use ChatGPT in immigration practice at all?
Yes, when used as a drafting accelerator under attorney supervision, with confidentiality protections in place, and with independent verification of any legal citation or factual claim. Comment 8 to ABA Rule 1.1 actually points toward AI competence as part of ongoing lawyer competence. The unethical path is not using AI — it is using it without knowing how it fails.
Can I paste redacted client facts into free ChatGPT?
No. Redaction is harder than it looks — a unique fact pattern can still identify a client. Free-tier prompts may be used for training. Use an enterprise account with zero-retention terms for anything that touches a real matter, or use fully synthetic facts on free tiers for prompt development only.
Which AI tool is best for reviewing long USCIS records?
Claude Sonnet or Opus with a 200k+ context window handles long records well. Gemini's million-token context is stronger for very large document dumps. ChatGPT with GPT-5-tier long context is competitive up to 128k. Whatever tool you pick, do not trust its summary of legally significant portions without reading them yourself.
What happens if ChatGPT invents a case citation in my filing?
Sanctions, potential bar discipline, embarrassment, and case damage. Multiple courts have already sanctioned attorneys for AI-hallucinated citations. Every citation must be independently verified. Better: have the model produce structural scaffolding only, and add citations yourself from a real database.
Should paralegals be allowed to run these prompts unsupervised?
They can run the prompts. They cannot send the outputs to clients or file them without an admitted attorney's review. Draft the internal policy that way explicitly. UPL exposure runs through the firm, not just the individual.
Do these prompts work for Canadian or UK immigration?
The structural scaffolding transfers. The regulatory content does not. Rewrite each prompt with the correct jurisdiction, forms, and terminology, and have counsel in that jurisdiction review before use. Do not assume US-trained model outputs map onto IRCC or UKVI practice.
How often should we update our prompt library?
Quarterly at minimum. Immediately after any USCIS form edition change, major policy memo, or executive order affecting your practice area. Assign one attorney as library owner with the authority to freeze outdated prompts.
Can ChatGPT help clients directly through a chatbot on our website?
Cautiously. A tightly scoped intake chatbot that only collects contact info and case type is fine. Anything that answers legal questions crosses into UPL and advice-without-engagement risk. Consult your state bar's guidance on AI-powered client-facing tools before deploying.
What is the single biggest mistake firms make with AI in immigration work?
Skipping attorney review because the output looked good. The outputs always look good. That is what the model is trained to do. The mistake is treating polished prose as verified content. Build the review step into the workflow so no one can skip it, even on a bad day.
Where do I start if my firm has never used AI before?
Pick one prompt from this article. Run it on ten past matters where you already know the right answer. Compare. Adjust. Then pilot it with one attorney and one paralegal for two weeks. Only then roll out firm-wide. The PromptSpace library has starter templates you can adapt.
Immigration lawyers who treat ChatGPT as a fast junior — useful, careful review required, never trusted with the final word — pull real hours back into their week. Immigration lawyers who treat it as a shortcut get sanctioned. Pick the first path, build the library, train the team, and keep your bar card intact.
