How PlainMedicare is made
Every figure on PlainMedicare comes from a public file of the Centers for Medicare & Medicaid Services (CMS). Each section names its file and date, and plan pages link Medicare Plan Finder, where you can confirm a plan before you enroll.
Which CMS files does it use?
- Plans, premiums and counties
- The CMS Medicare Advantage and Part D landscape files for 2027 (released September 22, 2026, revised September 30) and for 2026 (the September 2, 2026 revision). They list every plan sold to individuals, the counties where it is sold, its monthly premium, deductible and yearly out-of-pocket limit.
- What happens to each plan
- The CMS plan crosswalk from 2026 to 2027 (October 1, 2026; CMS crosswalk file). It says, for each 2026 plan, whether it renews, renews in fewer or more counties, is consolidated into another plan, or is terminated, and which 2027 plan its members move to.
- Members
- The CMS monthly enrollment reports by plan and by county for September 2026 (CMS enrollment data). CMS replaces any count under 11 with an asterisk to protect privacy; PlainMedicare shows those as “fewer than 11” and leaves them out of totals.
- Medicare Advantage share
- The CMS state and county Medicare Advantage penetration file for September 2026: people eligible for Medicare and people enrolled in a plan.
- Benefits and copays
- The CMS plan benefit package files for 2027 and 2026: the Part B premium reduction (the giveback), dental, vision, hearing and over-the-counter allowances, copays for doctor and specialist visits, emergency and urgent care, and the drug tiers.
- Star ratings
- The CMS 2026 Star Ratings, published in October 2025 (CMS Part C and D performance data). The 2027 ratings are due in October 2026 and will replace them when CMS publishes them.
- Drugs
- The CMS Part D formulary file for the second quarter of 2026 (which plans cover a drug, on which tier) and the CMS Medicare Part D Spending by Drug data for 2020 to 2024 (what Medicare paid for it and how many people used it).
How are the files joined?
A plan is identified by its contract number and plan number, for example H5216-393; a plan filed in segments (parts of its area with different benefits) is shown as one plan: its premium and out-of-pocket limit show their range by county, and each benefit line shows the most generous segment, marked ◆ (“best in its area”), because some counties get less. Each landscape county is matched to its federal county code, so a plan's counties, a county's plans and the enrollment counts line up. The crosswalk then links each 2026 plan to its 2027 plan, which is how a plan page can put 2026 beside 2027 the way an Annual Notice of Change does, and how a county page can count the plans that leave it.
A plan is called “not offered in 2027” only when every crosswalk row for it carries the CMS status “Terminated/Non-renewed Contract”. The crosswalk lists 765 individual plans that way; one of them, H9047-013 (a Part B-only plan), is not in the 2026 landscape file, so the site counts 764. It “folds into another plan” when CMS consolidates it and moves its members, and it “leaves a county” when none of its 2027 successor plans is sold there. Employer and union group plans (plan numbers 800 and up) are not in the landscape files; their contract pages show only what the enrollment report and the crosswalk say about them.
Drug names in the formulary file are grouped by active ingredient and brand, so one page answers for a drug however a plan lists it, and the spending figures are matched to the same ingredient name.
What stops a rebuild?
The pipeline refuses to publish when a source file is missing, when a landscape county row cannot be matched to a county, when a 2026 plan has no crosswalk row, when any page fails to render from its data, or when a page that readers reached from a search engine would stop answering. In the current build all 3,238 counties matched and every 2026 plan has a crosswalk status. Before a release, every page type is drawn at phone, tablet and desktop widths and checked by script for broken layouts, and a sample of pages from each type is read against the files.
Who reviews the pages, and when?
Before each release, three figures on every page type (plan, contract, county, state, insurer and drug pages, and the analyses) are checked against the CMS file each one cites, and the pages are read in separate review passes: against Google's and Bing's published quality guidelines, for claims a plan, an insurer or a reader could dispute, and for any figure that reads differently on two pages. The studio runs these checks with AI tools and records each one; the last full review was for the release of October 7, 2026. Between releases, every data load runs the checks above, and a page type whose figures change is checked again before it is published. PlainMedicare publishes the CMS files as data, not advice: confirm a plan's costs and coverage in Medicare Plan Finder or with the plan before you enroll.
What do AI tools do, and what do they not do?
The studio that operates PlainMedicare (see About) uses AI tools to write the pipeline code, the page designs and the explanatory text of the guides, and reads that text against the CMS files and Medicare.gov before it is published. The sentences on plan, county, state, insurer and drug pages are written by fixed templates from that page's own numbers. No figure is produced by a model: every number is the result of a database query over the files above.
What do the CMS files leave out?
The files describe what plans filed with CMS, not how a plan treats its members. Premiums and benefits can still be corrected by CMS before January, and a plan's network of doctors and pharmacies is not in these files, so confirm a plan on Medicare Plan Finder or with the plan before you enroll. Membership counts are a September 2026 snapshot. Star ratings are the 2026 ratings until CMS publishes those for 2027. Drug costs depend on your pharmacy and on where you are in the year, which a tier does not show. Prior authorization and referral rules are filed service by service in the benefit package files and are not summarized on plan pages; the plan's Evidence of Coverage lists them.
When is it updated?
The 2027 pages were rebuilt from the files released between September 22 and October 1, 2026. When CMS publishes a revision of the landscape or crosswalk files, the 2027 Star Ratings, or a new month of enrollment, the pipeline is run again and the dates in the source lines change with it. What changed on the site is listed on the changelog, and fixed errors on the corrections page.
What do readers ask about the figures?
Why is a benefit marked ◆ (“best in its area”)?
Some plans file their benefits in parts (segments) for different counties. The plan page shows the most generous part and marks the line with ◆, because members in other counties may get less; the plan's Evidence of Coverage gives the amounts for your county.
Why does my plan's letter say something different?
Your Annual Notice of Change is the plan's own statement for your county and decides what applies to you. The pages follow the CMS files of September and October 2026; CMS can still correct a plan's filing before January, and the site loads each revision when it appears.
Why do some drug pages have no 2024 spending?
CMS's Part D Spending by Drug file names each drug by its brand or generic name, and each row goes to one page only: a brand row to its brand's page; a generic row to the page named exactly like it, or else to the page with the same active ingredient and salt, oral forms first (CMS's “Doxepin HCl” row is the oral capsule's, so it sits on the oral doxepin page, not the cream's). A row never moves to a different active ingredient. A drug new since 2024, or one whose row belongs to another form of it, shows “no 2024 match” rather than a borrowed figure.
Why do county member counts say “fewer than 11”?
CMS hides enrollment counts under 11 in its county files to protect privacy, so the counties of a plan add up to slightly less than its total.
Who answers for it?
Questions and reported errors go to hello@plainmedicare.com. Each report is checked against the CMS file it concerns; the editorial policy says what happens next.