The AI Review Export packages your plan and its projection into a single block of text designed to be pasted into a chatbot like Claude, ChatGPT, or Gemini for a "second opinion" review. The AI reads your plan, the year-by-year projection, the per-year tax breakdown, and any engine warnings, then responds with observations, sanity checks, and suggestions you can take back into the planner.

The page has two modes, chosen from an Export type toggle at the top:

• Review one plan (default) - the classic single-plan export described below.
• Compare plans - exports two of your saved plans side-by-side for an AI-assisted comparison. Pick a base plan and a comparison plan, choose Standard or Full detail, and click Copy comparison. The AI sees an executive comparison table (success rate, ending balances, lifetime taxes / withdrawals / RMDs / Roth conversions / healthcare costs, ending bucket balances with deltas), a curated list of material input differences, a year-by-year deltas table, an auto-generated list of notable difference years (conversion-only, RMD-only, large tax / withdrawal deltas, depletion divergence), and a compact snapshot of each plan. Full additionally appends the raw year-by-year tables and per-plan Roth conversion schedules. The prompt explicitly asks the AI to surface tradeoffs rather than recommend one plan.

Why this exists

A skilled human reviewer would scan your year-by-year projection looking for things like withdrawal-rate spikes, RMD onset bumps, post-spouse-death tax-bracket compression, an under-funded healthcare gap before Medicare, or a Roth strategy that wastes a bracket. Modern chatbots can do a lot of that scan in seconds when given the right context. The AI Review Export hands them that context in a structured, computer-readable form.

How to generate it

Go to AI Review in the planner sidebar. In Review one plan mode, click Copy to put the single-plan export and AI-review prompt on your clipboard. In Compare plans mode, pick the two plans and click Copy comparison. Then paste into a chatbot to start the review session.

For the clearest comparison, make sure both selected plans have current simulation results. Comparisons are most meaningful when both plans use the same Monte Carlo settings and fixed seed - the export's metadata block flags any mismatch and the AI is told to caveat the comparison accordingly.

The page also shows a read-only Preview card with the exact text the Copy button puts on your clipboard, so you can see what's being shared before you share it.

What the single-plan export contains

The export is plain text, organized into sections so the AI can navigate it:

• General information - filing status, birth years, retirement and plan-to ages, state.
• Financial accounts - per-account balances, APYs, cost basis, dividend / distribution assumptions, allocation / custom-return overrides, withdrawal priority, owner.
• Assumptions - returns, allocation, inflation, Monte Carlo parameters.
• Income, Social Security, pensions, windfalls, expenses, debts, real estate, ACA, Medicare, LTC, Roth strategy - the configured plan inputs in human-readable form.
• Milestones & Events - both user-created and engine-generated milestones.
• Planning results summary - success rate, worst/median/best end balances and average returns.
• Year-by-year plan summary - the full projection table: income, expenses, taxes, net flow, RMD, distributions, per-account balances, portfolio total.
• Per-year tax breakdown - Federal, State, FICA, Cap Gains, NIIT, RMD Tax, Roth Conv Tax, Tax on SS, taxable income, ACA MAGI.
• Engine warnings - anomalies the engine itself flagged (cash floor breached, SEPP balance warning, ACA overlap, etc.).
• Export warnings - reconciliation drift, where the documented breakdown doesn't add up to the headline total. When this fires it usually means there's something worth investigating.
• Cash reserve top-ups, Roth conversion audit trail, per-bucket transactions, plan-notes-vs-structured-inputs comparison - additional detail useful for in-depth review.

What the comparison export contains

• AI prompt + guardrails - asks the AI to surface tradeoffs across taxes, buckets, RMDs, healthcare / IRMAA / ACA, and notable years. Explicitly tells the AI not to recommend a specific plan and not to provide individualized financial, tax, legal, or investment advice.
• Comparison metadata - timestamp, both plan names, whether Monte Carlo settings match, fixed-seed status (same paths / different paths / not pinned), and a stale-results warning when either plan needs a fresh simulation.
• Executive comparison table - side-by-side values plus a delta column for success rate, worst/median/best ending balances, lifetime taxes, lifetime portfolio withdrawals, lifetime RMDs, lifetime Roth conversions, lifetime healthcare costs, and ending bucket balances.
• Material input differences - curated, high-level diff across 14 areas (profile / plan-to age, state, accounts, Social Security, pensions, annuities, other income, recurring expenses, one-time expenses, windfalls, healthcare, Roth conversions, return assumptions, inflation). Includes an "inputs believed unchanged" list when applicable.
• Year-by-year comparison table - one row per projected year with base, comparison, and delta on ending portfolio, taxes, withdrawals, Roth conversions, RMDs, and tax-deferred / tax-free balances. The Notes column flags structural events (RMD-only-on-one-side, depletion-only-on-one-side, etc.).
• Notable difference years - auto-generated, ranked list of years that stand out: conversion-only years, RMD start years, tax deltas above $5k, withdrawal deltas above $10k, and portfolio depletion divergence.
• Compact plan snapshots for both plans - one-line summaries of each plan's profile, assumptions, accounts, income, Social Security, pensions, annuities, healthcare, expenses, windfalls, Roth conversions, and planning results.
• Full comparison mode (optional) - additionally appends the raw year-by-year tables for both plans plus per-plan Roth conversion schedules.

What an AI Review is good at

• Spotting numbers that look off at a glance - a tax row that's $0 in a year with six-figure ordinary income, a withdrawal rate over 8%, a windfall that doesn't show up in the right account.
• Suggesting specific changes - "convert another $20k in 2032; you have headroom under the IRMAA threshold," "consider claiming SS at 67 instead of 70 given the spending gap."
• Identifying transitions worth understanding - the filing-status change after a spouse's death, the RMD-onset jump, the year ACA coverage ends and Medicare begins.
• Cross-checking against tax rules - flagging missing OBBB senior bonus years, expected SS taxability levels, or NIIT triggers.

What an AI Review is not

• Not a substitute for a CFP, CPA, or other licensed professional. Use the AI's response as a starting point for questions; don't treat it as definitive advice.
• Not a guarantee. Modern chatbots are wrong about specific numbers more often than you'd expect, especially arithmetic on long tables. Verify any number-based claim back against the planner.
• Not a market predictor. An AI can't see the future, and neither can the Monte Carlo simulation - both are tools for thinking, not forecasts.
• Not private. The chatbot you paste into has its own data-handling rules. Retirement Figures has no connection to whatever AI tool you use, so anything you choose to share with it is governed by their terms, not ours.

Suggested workflow

1. Generate the export from the AI Review page.
2. Paste into your preferred chatbot - the prompt asks for an honest review with specific suggestions.
3. Read the response with skepticism. Check the suggestions against the planner.
4. If a suggestion seems worth trying, apply it on the relevant input page, then look at how the dashboard's numbers move.
5. Re-export and re-run the review when you've made substantial changes.

The cycle is what makes this useful - one round of AI feedback is interesting; iterating until the AI no longer finds material concerns is where the value compounds.