i re-checked pricing on 76 tool comparisons. 15 of them were wrong.

i run a small site that compares software tools and picks a winner. in june i wrote 69 of those comparisons. this month i went back and checked every price against the vendor's own pricing page.

fifteen were wrong. two were wrong enough that the verdict flipped — the tool i'd recommended was no longer the one i'd recommend.

that's a 20% error rate over roughly three months, on the single fact people come to a comparison page for. i don't think my site is unusual. i think this is what comparison content looks like everywhere, and nobody checks because checking is boring and nothing breaks visibly when you don't.

here's what actually changed.

the two that flipped a verdict

shortwave went from $9/month to $30/seat, and deleted its free tier.

my page said "shortwave for ai-native email at a fraction of the price" and recommended it over superhuman. that was true when i wrote it. it isn't now: shortwave's cheapest plan is business at $30/seat/month, then premier at $45 and max at $120. the free personal tier is gone, replaced by a 14-day trial.

meanwhile superhuman — now owned by grammarly — is $30/month, or $25 on annual billing. so the cheaper option is superhuman, which is the exact opposite of what my page told people for three months. i flipped the winner.

screen studio retired its one-time licence.

this one hurt more, because the entire argument of my screen-studio-vs-tella page was "pay once instead of subscribing." screen studio was famous for a ~$89 perpetual licence.

new buyers can't get it. it's subscription-only now: $20/month, or $9/month billed annually. existing lifetime holders keep theirs. i verified this directly on their site because i didn't believe the aggregator that told me.

the comparison still exists, but it's now subscription vs subscription, and the sentence that justified the whole page was dead.

the one where i'd been publishing something false

gumroad has been a full merchant of record since january 2025.

my gumroad vs lemon squeezy page said gumroad "leaves tax handling to you." it hasn't for over eighteen months — gumroad calculates, collects and remits sales tax, vat and gst worldwide, the same as lemon squeezy does.

that wasn't a stale price. that was me telling people something untrue about a tax obligation. gumroad's fee is also 10% + $0.50, plus card processing on top, which my page also had wrong.

the structural repricings

these are the ones that change how you get billed, not just how much — and they're the ones stale pages miss most often, because the headline number sometimes doesn't move at all.

• polar replaced one flat rate with four plans in may 2026. its free tier is now 5% + $0.50 — identical to lemon squeezy, which it used to undercut. the cheaper rates moved behind $20–$400/month plans. every article claiming polar is cheaper than lemon squeezy by default is now wrong.
• netlify dropped per-seat billing on pro in april 2026. $20/month now covers unlimited team members. vercel still charges $20 per deploying seat. on a team of five that's $20 against $100 — a structural advantage that didn't exist six months ago.
• framer went the other way and added a $20/month per-editor seat charge on every paid plan. a solo site that used to cost $5 now starts nearer $30.
• notion ai stopped being an $8/seat add-on. full ai now requires the $20 business plan, so "notion with ai" is $20/member, not $10 + $8.
• clerk raised its free tier from 10,000 to 50,000 monthly retained users in february. a lot of products that were about to start paying clerk no longer will.
• mixpanel cut its free tier from 20m events to 1m in late 2025. that erased the main advantage my posthog vs mixpanel page was built on.
• gsap became completely free in april 2025 after webflow acquired it, including every plugin that used to sit behind a club membership. my page still described morphsvg and drawsvg as paid.
• 1password added $12/year in march. bitwarden doubled premium from $9.99 to $19.80/year in january — its first increase in ten years.
• kit raised creator from $25 to $39. slack business+ went from $12.50 to $15.
• shadcn/ui switched its default primitive layer from radix to base ui in july 2026. not a price, but every tutorial written before then quietly assumes the wrong foundation.

what i got wrong about my sources

i started by trusting pricing aggregator sites. that was a mistake i had to correct twice.

one told me planetscale had brought back a free tier. planetscale's own pricing page says otherwise — there's still no free tier, and the cheapest route in is a $5/month single-node postgres instance. another two disagreed about unsplash+ at $9.99 and $20; the actual page says $12/month, or $4/month annually as a launch promo that unsplash explicitly warns may not renew at that rate.

the aggregators are mostly scraping each other. when two sources conflict, only the vendor's own page settles it.

the boring lesson

pricing pages are the most-read and least-maintained part of any comparison site. the incentive structure is bad: writing a new comparison gets you a new page that might rank, while re-checking an old one gets you nothing visible. so nobody does it, and the whole category slowly fills with confident, specific, wrong numbers.

i now stamp every price with the date it was verified and the source it came from. if a price on my site is stale, i'd rather it be obviously stale than quietly wrong.

if you maintain anything with prices in it, the 20% figure is worth sitting with. i'd assumed mine were mostly fine. they were mostly fine. "mostly" was doing a lot of work.

---

the comparisons are at whichtool.fyi, and the pricing on each one now shows when it was last checked. if you find a stale number, it's a bug and i'd like to know.