How a verdict gets decided
A label next to a price is worth nothing on its own. Here is the exact rule behind every YES, KINDA and NOT REALLY on this site — the four things we check, who decides, and where it goes wrong.
This archive asks one question of 1,358 tools: can AI replace it today? Not “will AI disrupt this category”, not “is this company doomed” — whether the specific job you are paying for has already moved into something you can use for free or already have.
Three answers, and a published rule for each. The rule is applied by a human, the reasoning is printed on the tool's page, and every entry carries a link to argue with it.
The three answers
given when A named, shipped AI feature does the core job this tool was paid for, and there is a free or already-bundled way to get that job done.
what to do Cancel it, and use the replacement listed on the entry.
given when The headline feature is now bundled elsewhere, but the workflow, data, integrations or compliance around it are still doing real work.
what to do Downgrade or re-scope it. Stop paying for the AI part; keep it if you use the rest.
given when A model overlaps with the tool but does not finish the job. The moat — workflow, proprietary data, taste or trust — held.
what to do Keep paying. Re-check when the named company ships in this category again.
The four things we check
These are the rows in the scorecard on every tool page, in the order they appear there.
- 01
A named launch, not a vibe
Every entry has to name the specific feature or product that did the damage, the company that shipped it, and the month it landed. “AI got better” is not a verdict. If we cannot name the launch, the tool does not go on the list.
limit Where an announcement link is recorded, the entry links it. Not every entry has one yet — the scorecard says which.
- 02
How much of the job it covers
The only question that matters: does the native feature do the job you were paying for, the headline part of it, or just something adjacent? Full coverage is YES, headline-only is KINDA, adjacent is NOT REALLY.
limit This is the judgement call, and it is marked “editorial call” on every tool page. It is where reasonable people will disagree with us.
- 03
Is there a free way to do it?
A verdict that leaves you nowhere to go is useless. Every entry lists where you would actually go instead, and flags which of those have a tier that covers the core job without a card.
limit Free tiers move. Rate limits tighten. Check before you cancel anything.
- 04
What the call is worth
The advertised entry-level paid tier, in USD per month, annualised on the page so you can see the yearly number you are deciding about. One-time and annual licences are stored as their monthly equivalent.
limit It is a comparison anchor, not a quote. Enterprise pricing is not tracked at all.
A worked example
The same four checks, applied to Frase.
YES — ChatGPT Search + deep research turned SERP briefs into a single prompt. ChatGPT and Perplexity now cover the same core job on a free tier, so the $45/mo is optional.
- A named launch, not a vibe
- ChatGPT Search + deep research turned SERP briefs into a single prompt. OpenAI, February 2, 2025 — no announcement link recorded yet.
- How much of the job it covers
- The whole core job — the thing the subscription was actually for.
- Is there a free way to do it?
- 2 of 3 listed replacements have a usable free tier: ChatGPT and Perplexity.
- What the call is worth
- $45/mo at the entry paid tier — $540 a year per seat.
Where the one-line reasons come from
Being straight about this matters more than sounding authoritative. Each tool's reason is one of two things, and the tool page says which at the bottom of the scorecard:
The default. The sentence is assembled from what that entry already records — the named launch, its date, the replacements listed, whether any is free, and the price. Nothing is added that the entry does not already claim, so the reason and the data can never drift apart.
1,358 of 1,358 entries
Where the composed sentence misses the point — a tool whose moat is something the fields do not capture — a human writes the reason instead, and it overrides the composed one. These are the entries we are slowly expanding.
0 of 1,358 entries
How complete the evidence actually is
Counted across the archive at build time, not estimated. Every tool page shows which of the three it has, so you can see how much to trust that particular verdict.
- Announcement linked
- 0%
2 / 1,358
A link to the launch that backs the verdict.
- Free replacement listed
- 85%
1,159 / 1,358
At least one escape hatch with a usable free tier.
- Two or more escape hatches
- 99%
1,342 / 1,358
Somewhere to go that is not just the first result.
Announcement coverage is the weakest of the three and we are not going to pretend otherwise. It is the thing being worked on.
What does not affect a verdict
- No vendor pays for a verdict, a rating, or a place on the list. Sponsorship exists, it is confined to the labelled slots in the side rails, and it buys nothing on this page or any tool page.
- No affiliate links. The replacements listed on an entry are plain links to the product.
- Votes do not set verdicts. They tell us which entries people care about, and which ones to re-check first.
- A company being on the list is not a prediction about the business. Plenty of tools rated YES have profitable enterprise revenue that will outlive the consumer pitch that died.
Where this is wrong
Prices move constantly and this archive lags them, so treat every dollar figure as indicative. Free tiers get rate-limited without an announcement. Coverage — criterion 02 — is a judgement about a category made by a person, and 177 entries carry the harshest reading of it. And an archive this size is uneven: an entry someone checked last week sits next to one nobody has looked at since it was written.
None of that is fixable by writing more confidently. It is fixable by being corrected, so the correction path is on every single tool page, next to the verdict it disputes.