Instead vs Cranston vs Magnetic: AI Tax Prep Cost 2027

As of 3 November 2026, the difference between Instead, Cranston and Magnetic is who does the work. Cranston charges a published $40 per prepared return and keys it into your tax software, Magnetic returns AI drafts reviewed by a licensed EA or CPA for credits, and Instead is a free tax platform that bills AI agent credits from $30.

TL;DR:

Cranston publishes $40 per return, so 300 returns cost $12,000 a season; its SOC 2 Type II examination is still in progress.

Magnetic sells credits (1 for a simple return, up to 3 for a complex one), delivers in 2 to 4 days, and says client data stays in the US and is never used to train models.

Instead's platform and e-file are free, but agent credits start at $30 each and the vendor does not say how many a return uses, so 300 returns could cost $9,000 to $27,000.

The IRS Office of Professional Responsibility said in June 2026 that Circular 230 duties apply in full to AI output, whichever tool prepares the draft.

Pilot each tool on 20 returns you have already filed, and compare cost per accepted return, not the licence fee.

What is the difference between Instead, Cranston and Magnetic?
How much do they cost a firm preparing 300 returns?
Where is client data held, and is it used to train models?
Which tool suits which kind of firm?
What do IRS rules require when a firm uses any of them?
How should a firm pilot them before the 2027 season?
Why does independent evidence matter when choosing AI tax prep software?
Realistic expectations: review time is the real cost
Find your next AI tax prep tool through The AI Ledger
Frequently asked questions

What is the difference between Instead, Cranston and Magnetic?

The three tools sit at different points on the same spectrum. Cranston is software your reviewer supervises, Magnetic is closer to an outsourced preparer, and Instead is a replacement for the tax engine itself.

Cranston reads client documents, compares them with last year's return, and enters the return in your own tax software. Its site names Drake Tax, CCH Axcess Tax, UltraTax CS, Lacerte and ProSeries. Nothing is posted or filed until your reviewer approves it, and you e-file from your own software on your own EFIN.

Magnetic is a 1040 service. Per its own site, the AI drafts the return, a licensed US-based EA or CPA checks it, and an upload-ready file comes back in 2 to 4 days on average. The directory listing describes Magnetic as working inside UltraTax, Drake and ProConnect, while the site now describes a file you import. Confirm the exact hand-off on your demo.

Instead is a full platform. It says it is IRS-approved to e-file 1040, 1120, 1120-S, 1065 and 1041 returns, with one agent per client that can work across linked entities. Its homepage still marks preparation and filing as an "imminent release", so check what is live in your account today.

Tool	Best for	Editor score	Price signal	Works with
Instead	Firms willing to move their tax engine	4.5	Free platform; credits from $30	Its own platform
Cranston	Flat per-return pricing on 1040 and business returns	4.2	$40 per prepared return	Drake, CCH Axcess, UltraTax CS, Lacerte, ProSeries
Magnetic	1040 overflow with human review included	4.2	Credits; rate on request	Import file for your tax software

Scores are editor scores from The AI Ledger directory, each verified against the vendor's site on its last verified date: Instead on 24 October 2026, Cranston on 29 October and Magnetic on 14 October.

Pro Tip: Ask each vendor which forms and states it prepares without a person touching them. "Handles 1040s" can mean a single W-2 filer or a return with three K-1s and rental schedules.

How much do they cost a firm preparing 300 returns?

Only Cranston can be priced exactly today: 300 returns at $40 is $12,000. Magnetic and Instead publish a unit (a credit) but not the full price, so the honest answer for them is a range and a break-even.

Cranston's pricing page states one rate for any form, with state returns included and no seat, platform or onboarding fees. A client who files both a personal and a business return is billed twice. Past 2,000 returns a season the firm negotiates volume pricing.

Magnetic's FAQ says simple returns cost 1 credit and complex returns up to 3, with the credit price shown on a demo. Instead's pricing page says credits are consumed according to complexity, shows an estimate before each workflow runs, and expire 12 months after purchase. Its homepage gives "starting at $30" per credit, with an annual agreement required.

Returns a season	Cranston at $40	Instead at 1, 2 or 3 credits per return, $30 each	Magnetic credits needed (80% simple, 20% at 3 credits)
100	$4,000	$3,000 / $6,000 / $9,000	140
300	$12,000	$9,000 / $18,000 / $27,000	420
1,000	$40,000	$30,000 / $60,000 / $90,000	1,400

The sums: Cranston is returns times $40. Instead is returns times credits per return times $30. For Magnetic, 300 returns at 80% simple is 240 returns at 1 credit plus 60 at 3 credits, which is 240 + 180 = 420 credits. The credits-per-return figures for Instead are illustrative, because the vendor publishes none, and volume discounts would lower the $30.

The break-even is the useful number. Magnetic's 420 credits cost the same as Cranston's $12,000 if a credit is priced at $28.57, since $12,000 divided by 420 is $28.57. Instead matches Cranston at $40 a return if the agent uses 1.33 credits per return at $30.

Three costs sit outside any licence. Your tax engine still bills you unless you move to Instead. Reviewer hours remain. And unused credits expire, so buy for a realistic volume.

Pro Tip: Ask Magnetic for the credit price at your volume and Instead for the credits used by a median return of yours. With those two numbers, the table above becomes your real comparison.

Where is client data held, and is it used to train models?

Magnetic is the only one of the three that answers both questions on its own pages. It says client data is stored and processed only in the US, with AES-256 encryption both in transit and at rest, and is never used to train models.

Magnetic's security page also says third-party models run in US-based, zero-data-retention deployments, and lists a written information security plan, a data processing addendum and an AI policy.

Cranston's security page says each firm has its own workspace, data is never shared across customers, and encryption is AES-256 at rest and TLS in transit, with role-based access, SSO and audit logs. It says its SOC 2 Type II examination is in progress. It does not state a hosting country or whether client data trains models on the pages read.

Instead links a security page described as covering SOC 2, compliance and trust. The address the AI Ledger tried returned a not-found page on 3 November 2026, and the pages read do not state a hosting location or a training policy. The directory notes that Instead states SOC 2 compliance.

Tool	Hosting stated	Trains on client data	Reads only or writes back	Assurance stated
Instead	Not stated	Not stated	Own platform, prepares and files	SOC 2 (vendor statement)
Cranston	Not stated	Not stated	Writes the return into your tax software; you file	SOC 2 Type II in progress
Magnetic	US only	No (vendor says never)	Delivers an import file after human review	WISP, DPA and AI policy published

"Not stated" means the vendor's pages did not say, which is different from a no. Put those questions in writing before loading live client data.

Which tool suits which kind of firm?

Choose by how much of the work you want to hand over and how much change you can absorb. The decision rules below assume a US firm filing mostly 1040s with some business returns.

Choose Cranston if you want predictable cost, keep Drake, CCH, UltraTax, Lacerte or ProSeries, and want your own staff to stay the reviewer. At 300 returns the bill is fixed at $12,000, and business returns cost the same $40.
Choose Magnetic if the bottleneck is preparer hours on individual returns and you would rather buy capacity than software. The EA or CPA review is included, and the vendor's error credit applies if a misentered or omitted figure changes liability by more than $10 (within 60 days, capped at the prep fee charged).
Choose Instead if you are willing to pilot a new core platform. It is the only one of the three that could replace your engine and its licence fees, which is a bigger change and a bigger saving.

Where each falls short. Cranston is very young and its SOC 2 Type II is unfinished. Magnetic's website is thin on detail, its guarantee covers the fee rather than your liability, and turnaround of 2 to 4 days matters in March. Instead's credit use per return is unpublished and some filing features are still rolling out.

Other tools overlap. Filed and Juno feed returns into Drake, Lacerte or CCH, and Black Ore Tax Autopilot prepares 1040s end to end. See the wider price comparison and the head-to-head comparison tool.

Pro Tip: If you run CCH Axcess Tax, ask Wolters Kluwer what is already in your licence before paying a third party for the same job.

What do IRS rules require when a firm uses any of them?

The firm stays responsible for the return. The IRS Office of Professional Responsibility said in a bulletin reported by The Tax Adviser that practitioners keep full responsibility for AI-assisted work and cannot rely solely on it.

The same bulletin tells firm leaders to write policies covering training, how client data is handled, monitoring of accuracy, and checks on any outside AI vendor. It also warns that charging for time not actually spent could raise an "unconscionable fee" concern, so efficiency gains should reach client billing.

IRS Publication 4557 adds that tax return preparers must create security plans under the FTC Safeguards Rule, and that firms must select service providers that maintain appropriate safeguards and require them by contract. An AI vendor is a service provider. Ask for its written information security plan and sub-processor list.

Adoption is already mainstream. In the 2026 survey reported by The Tax Adviser, 65% of 1,808 AICPA member respondents used AI for tax research, and only 16% had no plans to use it, down from 24.3% a year earlier. On engines, UltraTax CS leads at 22.3%, then Lacerte at 16.7%, Drake Tax at 14.3% and CCH Axcess Tax at 13.8%, which is why integration lists matter.

How should a firm pilot them before the 2027 season?

Run each tool on the same 20 returns you filed last season, with the answer known. Score accuracy, review time and total cost before touching live clients.

Pick 20 returns across your mix: simple W-2 filers, Schedule C, rental and at least three with K-1s.
Redact or get consent as your engagement letters require, and record where each vendor processes the data.
Time the reviewer from receiving the draft to sign-off, and log every correction by type.
Compare to filed returns line by line. Count misentered, omitted and misjudged items separately.
Compute cost per accepted return: licence or credits plus reviewer hours at your loaded rate, divided by returns accepted without rework.
Test the failure path: send corrected documents after the draft and see what it costs.

A tool that saves four preparer hours but adds two review hours has saved two, not four.

Why does independent evidence matter when choosing AI tax prep software?

Vendors report accuracy and time saved on their own terms, and a founder's testimonial is not a controlled test. Independent editor scores, dated verification and a consistent set of questions make tools comparable.

The AI Ledger does not test tools hands-on. Each listing is verified against the vendor's site and pricing pages on its last verified date, and scores are never influenced by paid placement. Browse every listing in the full directory.

Realistic expectations: review time is the real cost

Expect the first season to move work, not remove it. Budget for a full review of every AI draft, and treat any vendor accuracy percentage as a claim to test on your own returns.

Find your next AI tax prep tool through The AI Ledger

Start with the tax prep category to see every listed tool with its editor score, price signal and last verified date. Then use the finder to narrow by firm size and tax software.

Frequently asked questions

Is Instead cheaper than Cranston?

Not necessarily. Cranston charges $40 per return. Instead's platform and e-file are free, but agent credits start at $30 each under an annual agreement and a return may use several. At 1.33 credits per return Instead matches Cranston's $40, and above that it costs more. Instead may still save money by replacing your tax engine licence.

Does Magnetic use client data to train AI models?

Magnetic says it does not. Its security page states that taxpayer data is used only to prepare returns, never to train its models or anyone else's, and that third-party models run in US-based, zero-data-retention deployments. This is the vendor's own statement. Ask for the data processing addendum and sub-processor list to confirm it.

Is Cranston SOC 2 certified?

Not yet, according to Cranston. Its security page says the SOC 2 Type II examination is in progress, while encryption, role-based access, audit logs and SSO are live. A Type II report covers controls operating over a period, so ask for the timeline and the auditor's name before loading live client data.

How long does Magnetic take to return a 1040?

Magnetic says the average turnaround is 2 to 4 days from upload to an upload-ready file. Every return is checked by a licensed US-based EA or CPA before delivery. For a March crunch, test turnaround on your own volume, because an average hides the peak-week queue.

Can a CPA firm rely on AI-prepared returns without review?

No. The IRS Office of Professional Responsibility said in June 2026 that practitioners keep full responsibility for AI-assisted work and must verify facts, citations and calculations. AI output is a starting point. Under Circular 230, a qualified professional must make the final decisions, whichever tool prepared the draft.