Docketwise's AI writing assistant doesn't fill forms; its Smart Forms and document reader do, so Docketwise wins on the form itself. For letters it is closer: the assistant edits, summarises and translates inside the case file, while ChatGPT drafts more freely from redacted facts you paste in. Run one file through both before you pay.
The wrinkle is naming. Docketwise's pricing page lists "DocketWise IQ" while its writing-assistant and document-assistant pages say "8am IQ", and the pieces do different jobs: a writing assistant, a document assistant that reads passports and cards, and Smart Forms, which come on every plan with or without AI. The lowest list price is $80 per user per month, and the AI pieces arrive with Pro. I haven't run your files, so every result below is illustrative; the prices are the vendors' list prices.
What each tool is actually connected to
Plumbing decides what "filling a form" can mean. Docketwise's Smart Forms store a client's answers once and push them into whichever forms the case needs: you send one intake questionnaire, the responses populate the right areas of each form, and a bulk update changes a shared field across several forms in one pass. The document assistant (its pages also call it Legal Document AI or Data Capture) reads uploaded passports, residency cards, work-permit cards and arrival records, scans included, and drops names, birth dates and document numbers into Smart Forms and the client profile. The writing assistant is a separate piece. It edits grammar, tone and clarity, summarises notes, and translates between English and Spanish inside notes, emails, Smart Form invitations, comments and invoices.
ChatGPT has no such connection. It sees only what you paste or upload into that chat. It can't open your case, can't write into a form field and knows nothing about your firm until you tell it. What it offers is flexible reading and writing: a cover letter drafted from a fact sheet, a timeline compared against a list of documents, a client explanation reworded. Whatever it produces, you key into the form yourself. If you want the stage-by-stage view of where each tool sits in a filing, the intake-to-filing walkthrough covers it; this tutorial stays on one form and one decision.
| Job | Docketwise, from its own pages | ChatGPT | Where I'd start |
|---|---|---|---|
| Moving client answers into form fields | Smart Forms: one questionnaire fills several forms (all plans) | No link to the form; you re-key every answer | Docketwise |
| Reading passport and card data from scans | Document assistant (Pro) fills Smart Forms and the profile | Can read an uploaded scan, but that puts the passport image in a chat | Docketwise, then compare to the original |
| Tidying a cover letter you already wrote | Writing assistant (Pro): grammar, tone, clarity | Yes, from pasted text | Either |
| First-draft letter or declaration from facts | Pages mention drafting tools but give no detail | Yes, freely, from redacted facts | ChatGPT, after redaction |
| Spanish-language client updates | English and Spanish only | Many languages; quality varies by language | Docketwise for Spanish |
| Second read of a finished packet for mismatches | Form checking isn't described on the pages I read | Yes, from redacted text | ChatGPT as a second pair of eyes |
| E-filing and receipts | e-Filing is included from Pro | None | Docketwise |
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowSetting up a one-form test that is fair to both
Pick a form you file often, with a data-heavy first half and a short narrative part, such as an employment authorisation renewal. Build the test file from a matter you have already finished, so you know the right answers, but swap every identifier for an invented one: fictional name, made-up numbers in the same shape as real ones. That way no live client's data goes anywhere near a consumer-plan tool, and you can run the test as many times as you like. Docketwise's own product summary mentions a 7-day free trial, which is enough for this test if you build the dummy file before you sign up; ask whether the trial includes the Pro-level AI pieces, because the writing and document assistants come with Pro. For a longer evaluation of any legal AI app, the two-week trial scorecard adds accuracy, privacy and cost columns; the one-form test is the half-day version.
Seed the file with eight errors, written down before you start so you aren't scoring from memory.
| Seeded error | Where it hides | What should catch it |
|---|---|---|
| Two passport digits transposed | Questionnaire against the passport scan | You, comparing with the scan; the document assistant's extraction if it disagrees |
| Birth date with day and month swapped | Questionnaire against the scan | You; a second read if both dates are in the text |
| Middle name spelled two ways | Passport against the earlier approval notice | A second read; you |
| Four-month gap in the address history | The address list | A timeline check; you |
| Category code that contradicts the basis in the cover letter | Form against letter | You. A model can flag the mismatch, not decide which side is right |
| Yes/No answer left on its default | The form | You |
| Receipt number with one wrong digit | Form against the notice | You, against the notice itself |
| Cover letter quoting last year's fee or form edition | The letter | You, against the current form instructions |
Then run three routes on identical copies of the file. Route A is Docketwise only: send the questionnaire to a test email address, let Smart Forms populate, run the document assistant on the dummy passport scan, and use the writing assistant on the cover letter. Route B is ChatGPT only: paste a redacted fact sheet, ask for a field-by-field answer sheet, key it into the blank form, then paste the finished form's text back for a second read. Route C is Docketwise first with ChatGPT as the second reader on redacted text. Start a stopwatch when you open the file and stop it at "ready for lawyer review".
A scoresheet filled in with illustrative numbers
Log five things per route: minutes to ready-for-review, seeded errors still present at review, new errors the tool introduced, false alarms it raised, and the number of items you had to finish by hand. The figures below are invented to show how to read the sheet. They are not measurements of either product, and your own file will give different ones.
| Route | Minutes to review | Seeded errors left (of 8) | New errors introduced | False alarms |
|---|---|---|---|---|
| A: Docketwise only | 35 | 6 | 1 | 0 |
| B: ChatGPT only | 58 | 5 | 2 | 3 |
| C: Docketwise, then ChatGPT second read | 52 | 2 | 1 | 2 |
Three readings come out of a sheet like this. First, autofill is quick but copies errors faithfully: a wrong digit typed into the questionnaire arrives on every form that uses that field, which is why Route A still has six errors at review. Second, hand-keying from a ChatGPT answer sheet is slowest and introduces its own slips, such as a date typed in the wrong order. Third, the combination costs 17 minutes more than Route A (52 against 35) and leaves four fewer errors. At a paralegal rate of $30 an hour, an illustrative $0.50 a minute, those 17 minutes cost $8.50, or about $2.13 for each error removed. Whether that is worth paying depends on what a rejected or delayed filing costs you, and only you know that figure.
A realistic failure to look for in Route B: asked to "check this form", a model quietly corrects a mismatch in its answer instead of reporting it, so the discrepancy disappears from view. The prompt further down forbids that.
Cost per filing at list prices
Both tools bill per user per month, so cost per filing is the monthly bill divided by the filings you finish. The firm in this example has three users. Docketwise shows two sets of list prices on its pricing page, $80, $100 and $120 per user per month for Basic, Pro and Advanced, and $90, $120 and $140; the lower set appears to be the yearly price, so check which one your invoice follows. The table uses the lower set. ChatGPT Plus is $20 a month per user, and ChatGPT Business is $20 per seat a month billed annually or $25 monthly, with two seats as a minimum.
| Setup for 3 users | Monthly bill | Per filing at 10 a month | At 25 a month | At 50 a month |
|---|---|---|---|---|
| Docketwise Basic (3 × $80) | $240 | $24.00 | $9.60 | $4.80 |
| Docketwise Pro (3 × $100) | $300 | $30.00 | $12.00 | $6.00 |
| ChatGPT Plus (3 × $20) | $60 | $6.00 | $2.40 | $1.20 |
| ChatGPT Business, annual rate (3 × $20) | $60 | $6.00 | $2.40 | $1.20 |
| Docketwise Pro plus ChatGPT Business | $360 | $36.00 | $14.40 | $7.20 |
Two quick sums keep this honest. The step from Basic to Pro is $20 per user a month on the lower price set ($30 on the higher), but Pro also adds e-filing, e-signatures, a CRM and texting, so you are not buying the AI on its own. And the tools are cheap next to the time they touch: at $0.50 a minute, Pro's $12 per filing at 25 filings is repaid by 24 minutes saved on each filing, while ChatGPT's $2.40 is repaid by under five. That is why the scoresheet matters more than the price list. For the plan-by-plan detail, the Docketwise pricing tutorial goes further, and the vendor's own pricing page is the place to confirm today's numbers.
Privacy settings to check before either tool sees a file
What Docketwise says, and what it doesn't
The security page describes AES-256 encryption for data at rest on AWS, TLS for data in transit, restricted employee access and two-factor sign-in on all plans. The AI pages say outputs are optional and user-reviewed, and that the built-in design avoids the risks of external AI platforms. That last point is the vendor's claim, not something I can verify. What the pages I read do not say: whether client data is used to train models, which provider's model sits behind the assistant, and how long extracted passport data is kept. Ask in writing before you switch the AI on for live matters. A short message does it:
Before we enable the AI features on live matters, please confirm in writing:
1. Is any client content, including extracted document data, used to train or improve models, yours or a third party's?
2. Which companies process our content when we use the writing and document assistants?
3. How long are prompts, outputs and extracted data kept, and can we delete them?
4. Can AI features be switched off per user or per matter?
5. Where is the data processed and stored?What to set in ChatGPT, and what to strip first
Consumer ChatGPT plans have a model-training switch in the data controls, and Temporary Chat is not used for training but may be kept for up to 30 days. Business plans don't train on your content by default, according to OpenAI's pricing page. Setting labels move around, so check the current screen rather than trusting a screenshot from last month. If you are weighing seats, ChatGPT Business for law firms covers cost and settings in more depth.
Whatever the plan, strip identifiers before pasting. The aim is a fact sheet a stranger couldn't connect to a person, that still lets a consistency check work. The anonymising walkthrough has more patterns; this is the version for a filing.
| Field | Send ChatGPT | Why |
|---|---|---|
| Full names | [CLIENT], [SPOUSE], [EMPLOYER] | Names are the main identifier and a check never needs them |
| Birth date | Age in years, or the date only if the check needs it | You type the exact date into the form yourself |
| Passport, receipt and case numbers | [PASSPORT], [RECEIPT], [CASE] | These never leave the case system |
| Address history | Months at each address, no street names | A gap check needs durations, not locations |
| Dates of events | Real dates, only for the events being checked | A date mismatch is what you are hunting |
A second-read prompt, and what a good answer looks like
This is the prompt I would use in Route B or C. It asks for a report, not a rewrite, and it tells the model to leave gaps alone.
You are checking a draft immigration filing for internal consistency.
I will paste (1) a redacted fact sheet and (2) the text of the completed form.
Do not guess missing facts and do not correct anything.
List every place where the form disagrees with the fact sheet, or where two answers on the form contradict each other.
Output a table: field | form says | fact sheet says | confidence (high, medium, low).
Then list anything the form asks about that the fact sheet does not cover.
Do not give legal advice.An illustrative answer, with what you would do about each line:
| Field | Form says | Fact sheet says | Confidence | What you do |
|---|---|---|---|---|
| Birth date | 04/03/1991 | 3 April 1991 | Medium | Probably a format difference; check the form's date convention and the passport |
| Other names used | None | Passport shows a second middle name | High | A real flag; add it or confirm with the client |
| Address history | No entry June to October | No address recorded for those months | High | Call the client; fill the gap |
| Eligibility category | Code differs from the basis in the letter | Basis stated as in the letter | High | Yours to decide: the model sees the mismatch, not which side is right |
Notice what the model did not do: it did not know the right category, and it hedged the date. A tool that answered every row with high confidence would worry me more than one that hedges.
Letters: where the writing assistant and ChatGPT differ
For a letter you have already written, either tool can tidy it, so the choice is convenience. The writing assistant works where you already are, inside a note or an email in the case, and keeps the text in the platform. ChatGPT is looser: it can restructure a letter, propose a different order for exhibits or turn a long declaration into a short summary for the client, but only from text you paste. The shape of a typical edit looks like this. I wrote both versions to show the pattern; neither is captured output.
| Rough note in the file | After a tidy-up |
|---|---|
| client same job new dept, need renewal, old card expires soon, attach copy of card + passport page + approval notice | I write on behalf of [CLIENT] to request renewal of employment authorisation. The current card expires on [DATE]. [CLIENT] remains employed by [EMPLOYER] in the same role, now within a different department. Enclosed are a copy of the current card, the passport photo page and the approval notice. |
What you would fix in either version: the expiry date, the wording of the role change (did the role change or only the department?), and the list of enclosures against what is actually in the folder. A tidy sentence can promise an enclosure the file doesn't contain. Letters that rest on evidence need a heavier hand, and the evidence-letters tutorial covers drafting and redaction for those.
Translation is the other split. Docketwise translates between English and Spanish only, although its questionnaires can be translated into seven named languages and more. ChatGPT covers many languages, with quality that varies by language. Either is fine for helping a client understand a message. A document in another language that goes into the filing usually needs a translator's certification under the agency's rules, so check yours before treating any AI translation as anything but client communication.
What stays manual whichever tool wins
- The legal call. Eligibility, category and strategy are yours. A model can point at a mismatch; it cannot tell you which answer is right.
- Source-document checks. Compare every date and number against the original document, not against the questionnaire the tool copied from.
- Form edition and fee. Confirm the current edition and fee on the agency's own instructions before the packet is signed.
- Sensitive questions. Ask the client about criminal history and similar answers live, and record what they said, rather than letting a tool fill from a document.
- Client review and signature. The client reads the final text, with an interpreter where needed, before anyone signs.
- E-filing receipts and deadlines. Enter each receipt and calendar every follow-up date yourself.
- A supervision note. Record who reviewed the packet and which tools touched it, so the file shows human judgement at each step.
Turning the result into a rule the whole firm follows
A scoresheet only helps if it ends in a written rule, because the next paralegal won't have seen the test. If I had to pick one starting point for a small firm, I would turn on Smart Forms and the document assistant first, since that is where the keying time sits, and add a redacted ChatGPT second read only for filings where an error is expensive. The scan reader deserves one standing check of its own: any tool that reads images can confuse similar characters, such as a zero and the letter O, so compare extracted passport and card numbers against the original every time. Here is a draft rule you could adapt after your own test.
Client data goes into the case system only. Smart Forms and the document assistant prepare the first draft of every form; the reviewer compares every extracted number and date with the original document. The writing assistant may tidy client emails and cover letters. ChatGPT may be used only on a redacted fact sheet, never on passports, receipts or case numbers, and its output is a list of questions for the reviewer, not text for the filing. Every packet records the reviewer, the tools used and the date of review.
Repeat the test whenever you add a form type or a vendor changes its AI features, and file each scoresheet with the procedure so the next review starts from evidence rather than memory.
Docketwise and ChatGPT: what lawyers ask next
Does Docketwise's AI fill the forms for me?
Not the writing assistant. Forms are filled by Smart Forms, which turn one client questionnaire into answers on several forms, and by the document assistant, which reads passports and cards into those fields. Smart Forms sit on every plan; the AI pieces come with Pro. Either way the fill is only as good as the data it copies, so you still compare against the source documents.
Can I paste a client's passport into ChatGPT?
I wouldn't on a free or Plus account, where the chat may be used to improve models unless you switch that off, and even on a business plan I'd rather type identifiers into the form yourself. Send ChatGPT a redacted fact sheet instead: placeholders for names, numbers and addresses, and only the dates a consistency check needs.
Is ChatGPT Business safe enough for client files?
OpenAI says Business content isn't used for training by default, which removes one worry. Whether it is safe enough for your files depends on your confidentiality duties, your engagement terms, who in the firm can see chats and how long they are kept. Check your regulator's guidance and your insurer, and redact anyway.
How long should the one-form test take?
Half a day for one form, three routes and a scoresheet, using a dummy client built from a finished matter. Repeat it once on a second form type before you decide, because a tool that copes with a data-heavy form can still stumble on a narrative-heavy one.
Further reads
- Can AI Help Immigration Law Firms Answer an RFE? What to Check — The same tools on a harder document: answering a request for evidence.
- Immigration Agent AI: Client Document Checklists and Case Updates — Client document checklists and status messages after the filing.
- Immigration Case Software With Automation: A Buyer's Checklist — What to ask any case system about automation before you commit.
- Free ChatGPT for Lawyers: Where the Free Plan Stops Being Safe — Where the free ChatGPT plan stops being safe for client work.
- Best AI Chatbot for Lawyers: General Models vs CoCounsel — General models against purpose-built legal assistants.
- Which Legal Tasks Should a Small Firm Never Hand to AI? — A firm-wide list of tasks that stay with a lawyer.
- Claude or ChatGPT for a Law Firm: Which One to Pay For — If you decide a general assistant earns a paid seat.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Docketwise pricing, Smart Forms, AI writing assistant, Legal Document AI and security pages; OpenAI's published ChatGPT plan pricing and data-controls information.