Methodology

This is how the Enshittification Index turns a documented record into a number. New to the term itself? Start with What is enshittification? Everything below is mechanical: there is no editorial thumb on the scale, and if the rubric produces a number we don’t like, the number stands. Currently: 42 case files, 310 documented symptoms.

What gets a company included

• A documented pattern, not an incident: at least 6 sourced symptoms spanning at least 2 symptom categories. One scandal is news; a pattern is enshittification.
• Every symptom must be a dated, specific decision or event: a price increase, an ad insertion, a removed feature, a privacy rollback, a lock-in mechanism. Not vibes, decline narratives, or general criticism.
• Every symptom must cite a qualifying source (see the publication tiers below). No source, no entry. Candidate companies whose records can’t clear this bar are left off; several have been.
• Errors are corrected, not defended. If an entry is wrong, it gets fixed or removed.

Symptom categories

Each symptom is classified into one of six categories. Categories carry different point caps in the severity formula, weighted by user harm. Privacy violations and lock-in outrank a removed feature:

• privacycap 26 pts
• lock-incap 26 pts
• pricingcap 21 pts
• adscap 21 pts
• qualitycap 21 pts
• featurescap 15 pts

How publications are graded & weighted

Not every link is evidence. Every citation on this site is graded into a tier, and the tier mechanically scales how much the event counts toward severity. The test is editorial accountability, not the domain: a staffed, edited publication hosted on Medium qualifies; a self-published post on a fancy custom domain does not. Where both exist, we cite the strongest available source, and a paywalled original beats a free secondhand summary.

Two special cases. A company’s own announcements are Tier A for claims about what the company did. Its blog post announcing a price increase is the company convicting itself. Advocacy organizations (EFF, Authors Guild) are citable for their own documented actions and findings, attributed as such, at Tier C weight.

The severity formula

Severity (0–100) measures the current condition, computed mechanically from the event record. Evidence ages on a half-life: a fresh symptom counts at full strength and its weight halves every 48 months. It never reaches zero on the timeline, it just stops dominating the number. Documented remedies, reversals that clear the same sourcing bar, subtract evidence:

for each category c:
  E  =  max(0, Σ over events in c of:
          direction            (decay +1, remedy −0.6)
        × source-tier weight   (A/B = 1.0, C = 0.6; strongest citation counts)
        × 1.5 × 0.5^(age in months / 48))

  points_c  =  cap_c × (1 − 0.45^E)

severity  =  min(99, Σ points_c)

• Diminishing returns: each added symptom in a category contributes less than the one before, so the scale can’t be saturated by repetition.
• Breadth matters: decaying on many fronts scores higher than many events on one front, which is the actual enshittification pattern.
• Time matters: evidence halves every 48 months. A company that stops generating symptoms genuinely recovers; one that keeps decaying holds its score by earning it, every year.
• Remedies count, at a discount. A documented reversal (a price cut, a restored feature, tracking removed) subtracts evidence in its category at 0.6× its source-tier weight. Discounted because reversals are often partial and announcements are cheap; it still takes the same qualifying source. Talk is not a remedy. A category’s evidence never goes below zero.
• One behavior, one event. Additional coverage of a symptom already on the record attaches to that event as a corroborating citation (the strongest source sets its weight) instead of becoming a new event. Repetition in the press is not repetition in the record.
• 99 is the ceiling for the living. A score of 100 is reserved for platforms that have actually been shut down, and almost none are. Enshittified platforms rarely die; they shamble on as zombies, which is rather the point.

Two derived readings appear alongside the score. Peak condition is the worst severity a record ever produced, replayed month by month under this same formula: the permanent rap sheet, displayed but never fed back into the score. A file is labeled in remission when its current severity sits at least 10 points below that peak. The trend arrow is the score’s change over the trailing 12 months.

The headline 💩 rating is severity ÷ 20, rounded: 💩 out of 5. More 💩, more documented decay. It measures the record, not the product as a whole; a low score may simply mean the file is young.

Condition

Severity maps to a condition label like a patient chart: a description of the documented record to date, not a forecast:

• TERMINAL80–100
• SEVERE60–79
• ACUTE40–59
• MODERATE20–39
• MILD0–19

Rubric changelog

v2 — 2026-07-07. The original formula counted evidence forever at full weight, so every score could only ratchet upward: given enough years of coverage, every company would end up TERMINAL and the index would stop discriminating. v2 replaces the binary recency bump (×1.5 inside 24 months) with the 48-month half-life above, adds remedy events, and adopts the one-behavior-one-event corroboration rule. Every score was recomputed under v2 on the same event records; nothing about any company’s record changed, only how time weighs it. The full reprice, for the record:

The Wall: house rules

Each case file has a Wall: anonymous, unverified patient testimony. It is venting, not evidence. Nothing on the Wall counts toward severity, and it is kept visibly separate from the sourced record. No account, no login, no tracking; a privacy- respecting bot check and a salted, truncated IP hash for rate-limiting only.

• Profanity is fine. Rage, all-caps, creative cursing. That’s the product.
• Slurs and hate speech are not. Filtered automatically. We document enshittification; we don’t host it.
• No threats. Possible threats are held for human review before anything publishes.
• No doxxing. Emails, phone numbers, and street addresses are filtered. Yell at companies, not people.
• No links. Kills spam dead. Cite sources in a symptom submission instead; that’s how something gets into the actual record.
• Reports work. Any comment reported three times is pulled pending review.

Independence & alternatives

This site is independent and is not affiliated with Cory Doctorow or his publisher. The alternatives on each case file are the options most commonly cited in coverage and user communities, listed as starting points, not endorsements. We have not independently tested them; do your own diligence. If a listed alternative has its own case file in this index, that is disclosed next to it. Some links may become affiliate links, which never changes what is listed or how it is described. For the theory behind the pattern, read the book: Enshittification: Why Everything Suddenly Got Worse and What to Do About It.