Plan vs. Bill: Upload Your Plan, Check the Bill, See the Truth
A deep dive on the "compare your insurance plan to your bill" concept — dental first, any insurance later. Real competition data, real keyword numbers, honest AI threat analysis.
Verdict
Build the system, not the app. The app alone is dead in 18 months. The system is a real business.
The app — upload two documents, get a comparison — is a feature. ChatGPT already does it. Every AI assistant will do it natively within a year. If the pitch deck says "AI compares your bill to your plan," that deck is worthless.
The system — code-explainer content that ranks + a shared database of parsed plan documents + the comparison tool as the entry point — is defensible. AI can compare two documents a user uploads. AI cannot answer "what does Delta Dental PPO actually pay for code D2740, and which upgrade clause is about to bite you" at scale, from fresh verified tables. That corpus is the business.
Verdict: build it Dental first Content + data, tool as lead magnet Do not raise money on the app
The Gap: What Exists vs. What Does Not
What the market has
• Bill negotiation services — humans argue the bill down after the fact (Goodbill, Resolve)
• Fair-price scanners — "is this charge normal compared to market rates?" (HealthScan AI)
• EOB explainers — plain-English translation of one document (DecodeMyForm AI)
• Professional coder tools — CPT/ICD lookup for billers, subscription-priced (Codify by AAPC)
• Denial help — fight the insurance company after they say no (Counterforce Health)
What nobody has
• Plan vs. bill, line by line — YOUR plan's actual copay table matched against YOUR treatment plan or bill
• Pre-treatment checking — before you sign the treatment plan, not after the bill arrives
• Plain-English code pages for consumers — every existing code site is built for professional coders
• The neutral tone — "here is what to ask about," not "they ripped you off, hire us"
• A shared library of plan documents — see the Plan Database section
Competition: Who Is Actually In This Space
| Player | What they do | Model | Compares to YOUR plan? | Threat level |
|---|---|---|---|---|
| Goodbill goodbill.com |
Reviews hospital bills for errors, negotiates reductions on your behalf | Human service, fee on savings | No — compares to fair pricing | LOW Different business (post-bill, high-dollar, human) |
| Resolve resolvemedicalbills.com |
Dedicated advocate audits and negotiates hospital bills, handles appeals | Human service, fee on savings | No — negotiation, not plan matching | LOW Could white-label or refer to you |
| HealthScan AI Android app |
Scans medical bills, translates CPT codes, compares charges against fair prices | Consumer app | No — fair-market comparison, not plan comparison | MEDIUM Closest neighbor; wrong reference point |
| DecodeMyForm AI | Explains confusing medical bills and EOBs in plain English | Consumer tool | No — explains one document, no matching | MEDIUM Half the feature |
| Counterforce Health | AI analyzes bills and records, helps understand and fight denials | Consumer service | No — denial focused | MEDIUM Downstream of your moment |
| Codify by AAPC / ICD10data | Code lookup and billing reference for professional coders | Subscription, B2B | No — code reference only | HIGH on SEO Owns the code-lookup keywords today |
| Cassidy AI cassidyai.com |
Compares health plan renewal quotes and proposals automatically | B2B (brokers, HR) | Yes, but employer/broker side, not consumer | LOW Validates B2B demand for plan data |
| ChatGPT / Gemini generic AI assistants |
Anyone can upload both documents and ask "does this match my plan?" | Free with existing subscription | Yes, if the user thinks to ask and does the work | HIGH This is the mow-over risk. See next section |
Read of the field: the industry organized around two moments — after the bill hurts (negotiation, denial fighting) and for professionals (code tools). The moment in the middle — "I have a treatment plan in hand, what will this actually cost me under my plan" — is unoccupied. That is the wedge.
Demand & Keywords: The Content Goldmine Is Real
Your instinct is correct and the numbers back it up. Consumer-facing explanation keywords have almost zero SEO difficulty, while professional code-lookup keywords are contested (owned by AAPC-class incumbents). The strategy: skip the head terms coders fight over, own the long tail consumers actually type.
| Keyword | Searches / month (US) | SEO difficulty (0-100) | Cost per click | Read |
|---|---|---|---|---|
| how to dispute a medical bill | 880 | 0 | $7.24 | Free to win, advertiser money in the clicks |
| dental crown cost with insurance | 1,900 | 0 | $6.93 | Your exact story. Untouched |
| how much does a crown cost with insurance | 2,400 | not yet contested | $5.75 | Same cluster, bigger volume |
| dental codes list | 390 | 0 | $2.28 | Gateway page to every code explainer |
| how to read an EOB | 260 | 5 | $0.35 | Easy win, top-of-funnel education |
| insurance explanation of benefits | 1,900 | 11 | $11.52 | Winnable with big advertiser value |
| how to appeal insurance claim denial | 210 | low | $16.67 | Lawyer money — small volume, premium clicks |
| surprise medical bill | 480 | low | $15.55 | News hook + policy content |
| how to negotiate hospital bill | 720 | low | $3.41 | Good content, moderate monetization |
| dental insurance waiting period | 480 | 40 | $8.19 | Getting competitive but valuable |
| annual maximum dental insurance | 320 | low | $4.31 | Explainer win |
| what is dental code D2740 | 50 | 2 | — | One of ~800 CDT codes. The long tail IS the play |
| HCPCS code lookup | 1,300 | 38 | $1.02 | Contested by incumbents |
| ICD 10 code lookup | 12,100 | 45 | $2.03 | Big volume, incumbent territory. Do not attack head-on |
| CPT code lookup | 6,600 | 50 | $2.53 | Same — Codify by AAPC defends this |
The aggregation math: a single code page is 20-80 searches a month — laughable alone. Eight hundred dental code pages at a conservative 25 average is 20,000 monthly searches of pure long tail, with no incumbent defending it because professional coders never search it and consumer sites never built it. Add plan-specific pages from the database (next section) and the addressable page count goes into the thousands.
Will AI Mow This Over in 6 Months?
Honest answer: the comparison feature, yes. The business, no — if you build the right layer. The CNET test from mid-2026 already showed ChatGPT finding medical bill errors from an upload. The generic "read two documents and diff them" capability is gone as a differentiator. Here is the threat timeline:
Now to 6 months — the feature commoditizes
Anyone who thinks to ask ChatGPT gets the comparison for free. Note the qualifier: anyone who thinks to ask. Most people never conceive of lining up the SOB against the bill — that awareness gap is the product's real job. Marketing, not technology.
6 to 12 months — AI assistants absorb the explainer surface
"What does code D2740 mean" gets answered inline by search engines and assistants, no click needed. Pure definition pages lose traffic. Defense: your code pages must carry what the model cannot — plan-specific numbers ("what you actually pay under the top 20 dental plans") and current-year verified tables. Definition content alone is sand; definition plus your database is concrete.
12 to 24 months — the plan database becomes the moat
Models can read any document a user uploads. Models cannot have a fresh, verified, structured corpus of thousands of real plan tables with usage signals. That asset appreciates while model capability depreciates around it. This is exactly why your shared-plan-database idea is the right pivot.
24 months plus — carriers move last, if ever
"Enter a code, see your copay" exists inside some carrier portals today and nobody finds it, because the incentive is opacity. Regulatory pressure (transparency-in-coverage machine-readable files) creates data but deliberately un-consumer-friendly surfaces. The gap survives on carrier incentives, not on technology.
The Plan Database: The Actual Moat
Your idea — a shared library of insurance plans, uploaded with permission, not tied to people, selectable on the next visit — is the correct answer to the AI problem. Here is the design as it should work:
Store the parsed table, not the document
Keep the extracted facts — carrier, plan name, plan year, tier, procedure code, coverage percentage, copay, downgrade clauses, frequency limits, waiting periods, annual maximum. Keep a link to the public source document. Facts are not copyrightable; wholesale republishing of carrier PDFs is the risk you avoid by never hosting the PDF itself.
Not tied to people = clean design
A copay table with no name attached is not personal health information. Plans are published benefit documents, not patient records. Users select their plan from the library; only the bill photo ever touches personal data, and that gets parsed and dropped. This is the HIPAA-clean architecture from the earlier discussion, formalized.
Permission is one checkbox
At upload: "Add this plan to the shared library so others can use it. It contains no information about you." Yes/No, default no, easy to change. Treat the plan document itself as semi-confidential anyway — some employer plans name the employer — so publishing is table extraction only.
The maintenance problem is the real cost
Plans renew every January. Stale data destroys trust — the one thing this product sells. Every table is versioned by plan year, stamped "verified for 2026," and freshness gets confirmed by usage: "did this match your bill? yes/no" is a free verification signal nobody else collects.
The flywheel this creates
Seed it yourself before any users arrive
The top 20 dental carriers' most common plans are publicly downloadable — carrier sites, employer handbooks, state filings. Parse those first. The library is never empty; cold start solved without waiting for uploads.
Every plan becomes a page
"Delta Dental PPO 2026: what it actually pays for crowns, implants, and the upgrade clauses" — programmatic pages backed by your tables. These rank for plan-specific queries no AI Overview can answer without your data.
Users select instead of upload
The tool gets faster and more accurate every time the library grows. Comparison quality becomes a data advantage, not a model advantage — models are interchangeable, your corpus is not.
The corpus gets sold, too
Brokers, HR platforms, dental offices writing treatment plans, TPAs — Cassidy AI already sells plan comparison to the broker side, which proves money exists for structured plan data. Year-two revenue, built on year-one uploads.
Revenue Math
Four paths, ordered by time-to-money. The first two fund the rest.
Content traffic + affiliate FASTEST
Code explainers and cost pages monetize via dental savings plan referrals, medical bill advocacy referrals, and insurance broker leads. Cost-per-click data ($5-16 on problem keywords) proves advertiser density in this exact traffic. Illustrative: 20,000 monthly visits at a conservative 2% referral rate and $15 per action is roughly $6,000/month — order-of-magnitude, verify offers before forecasting.
Display ads on reference pages
Code lookup traffic is high-frequency, low-duration — textbook display inventory. Dental and insurance advertisers pay premium rates. Adds a floor of a few hundred dollars per month at modest traffic.
Freemium tool
Free comparisons drive database growth and word of mouth. Paid tier at $7-10 per month: household plan vault, saved history, unlimited checks, renewal alerts when the plan year flips. Do not lead with the paywall — the free product is the marketing.
Plan database licensing (B2B)
The corpus sold to brokers, dental offices, HR tech. Cassidy AI's existence validates the spend. This is the asset-value play and only becomes sellable once the library is deep and verified.
Scorecard
| Dimension | Grade | Why |
|---|---|---|
| Market need | A | 30% first-pass denial rate, 32% of those from coding issues, everyone has a story. The need is not in question |
| Timing | A- | AI made the parsing cheap and the awareness possible; the window for owning the content surface is open now and will narrow |
| Scalability | A- | Content and data scale to zero marginal cost. Everything past the corpus is copyable; the corpus itself is the exception |
| Defensibility (app alone) | C | One prompt away from free. Honest grade |
| Defensibility (system) | B+ | Plan corpus + verification signals + audience. Not unassailable, but it compounds |
| Offer structure | B- | Indirect monetization to start (affiliate, ads). The paid tier and B2B licensing are where the offer gets strong |
| AI threat level | HIGH (feature) / MEDIUM (system) | The feature dies on schedule. The system lives if the database ships early |
Action Playbook
Phase 1 — Prove the content engine (Month 1-3)
Ship 50 dental code explainer pages from the real CDT list — each one: what the code means, what plans typically pay, the gotcha attached to it (alternate benefit, frequency limits, waiting periods). Ship 10 money pages targeting the KD-0 cost keywords ("dental crown cost with insurance"). Seed the plan database with the top 20 dental plans, parsed. Measure: rankings and clicks by month 3.
Phase 2 — Launch the tool on top (Month 3-6)
Comparison tool live: select your plan from the library (or upload), snap the bill or treatment plan, get the line-by-line output with the explain-before-you-accuse ladder. Permission checkbox feeds the database. Every comparison that runs on a library plan returns the "did this match" verification signal. Measure: comparisons run, library growth, upload-to-select ratio.
Phase 3 — Monetize (Month 6-12)
Affiliate offers on cost pages (dental savings plans, bill advocacy referrals), display on reference pages, paid tier for the household vault. Expand the content to medical codes and the auto/home explainers. Measure: revenue per thousand visits, paid conversion, retention on the vault.
Phase 4 — Sell the data (Year 2)
With a deep verified corpus: licensing conversations with brokers, dental practice software, HR benefits platforms. Measure: inbound interest from plan pages, licensing conversations started.