i checked the pricing on the big comparison sites. a lot of it is wrong.

i run a comparison site with 17,000 lifetime impressions. g2 and capterra have audiences i will not catch this decade. so this is not a takedown, and i am not pretending to be a competitor.

it is just that i spent the last two months re-checking pricing against vendors' own pages, and the same thing kept happening: the number everyone repeats is not the number on the vendor's site.

here are five, with sources.

the one that actually matters

bonsai is the interesting one, because nobody is lying.

bonsai's cheapest plan genuinely is $15 a month. every comparison site reports that correctly. the problem is that the $15 plan cannot send an invoice — basic covers time tracking, tasks, projects and CRM. invoicing and payments start at essentials, $25 a month, or $19 billed annually.

so when a page compares bonsai to an invoicing tool and prints "from $15", it is quoting a true number that answers the wrong question. the reader is choosing an invoicing tool. the number they need is $25. that is a 67% error produced entirely by accurate reporting.

this is the failure mode i would actually warn people about. not fabricated prices — correct prices attached to the wrong plan.

i found this one because i was writing quire vs bonsai, and quire is my own invoicing tool. so take the framing with the appropriate salt: i had a reason to look closely at bonsai's pricing. the $25 figure is on bonsai's own page either way, and the comparison discloses the conflict in its first line rather than its footer.

why it keeps happening

three reasons, none of them villainous.

aggregators cite each other. when two sites disagree about invoice simple by 40%, neither is reading the vendor's page, because the vendor returns a 403 to anything automated. they are reading each other, and the error gets laundered into consensus by repetition.

nobody is paid to re-check. writing a new comparison produces a new page that might rank. re-checking an old one produces nothing visible. the incentive points one way and it is not toward maintenance.

pricing changed faster than usual this year. cursor restructured in october — the $20 pro plan does not exist for new users. three weeks later, september-dated articles were still quoting it as current. that is not negligence so much as physics.

i am not clean either

worth stating plainly, because otherwise this reads as someone else's problem.

in september i re-verified 69 comparisons i had written in june. fifteen were wrong. two were wrong enough that the recommendation flipped — the tool i had told people to buy was no longer the one i would buy. one page said gumroad leaves tax handling to you, which had been untrue for eighteen months.

that is a 20% error rate over three months on the single fact people come to a comparison page for. my site is not better maintained than g2's by virtue of being smaller. it is better maintained because i went and checked, and i only know the rate because i measured it.

what i would actually tell a reader

not "use my site instead". three habits that work anywhere:

check the vendor's own pricing page. it takes forty seconds and it beats every secondary source including this one. if a comparison does not tell you when it last checked, assume it has not.

check which plan includes the feature you came for. the headline price is almost always the plan that cannot do the thing. this single check catches most of the real-world error.

discount any undated price. a number without a date is a claim about the past wearing present tense. this category reprices every few months.

what i changed on my own site

every price on whichtool now lives in a typed dataset with a mandatory verification date, rendered into a table on each comparison, with one rule: if a price is not verified at the vendor's page, the tool is left out rather than guessed. invoice simple is absent from a comparison it belongs in, for exactly that reason, and the page says so.

that is not a moat. anyone could do it, and the big sites could do it far better than i can with a fraction of the effort. it also does not exempt me: the one comparison on the site featuring a tool i built says so in the verdict, in a dedicated section and in the first question, and still hands the win to bonsai. it is just that at 17,000 impressions, being right is the only advantage available to me, and it happens to be the one readers actually wanted.

---

checked october 2026. every figure above was read from the vendor's own pricing page on the date shown, except invoice simple, which cannot be read. if one of these is stale by the time you get here, that is the whole point, and i would like to know.