Skip to content

AI Research

How AI Research works

TheMacApps uses AI to read public pages about Mac apps and compare them factor by factor, so you don't have to open dozens of tabs. This page explains where the facts come from, how a report reaches a result, and how to correct it when it's wrong.

No one can pay to change a result

Research pages have no sponsored placements. Payments, partnerships and votes are not inputs to the scoring formula below, and makers can't edit a report. A maker who spots a mistake uses the same correction process as everyone else, with a source.

How a report is built

  1. Start from a link or a list of apps. A quick preview reads the page you paste and finds comparable apps. It usually takes 10 to 20 seconds and is a first pass, not a verdict.
  2. Research each factor for each app. Full AI Research checks official sites, documentation, release notes, the App Store and public discussions. It is slower because it looks up every factor for every app.
  3. Keep the source with the answer. Every cell stores the page it came from and the date it was checked. Open any cell to see them.
  4. Score only what is confirmed. The formula below runs on confirmed answers only, so a missing fact can't quietly push an app down.

What each answer can be

Confirmed
At least one cited source supports the answer, and the AI's confidence in reading it is 70% or higher. Only confirmed answers can count toward a result.
Conflicting
Sources disagree. Both are shown and neither is picked for you.
Unknown
The AI looked and could not find a source that settles it. Unknown never counts as “no”.
Not researched yet
Nobody has looked yet. It stays open until someone does.

The scoring formula

Whoever creates a Research marks how much each factor matters. Each level has a fixed weight:

Must-have
5
Important
3
Normal
1
Low
0.5
  • Priorities decide. Only must-have and important factors are scored. If none are marked, the first five factors are used instead.
  • Points need a source. An app earns a factor's weight only when a confirmed source says it has that capability.
  • Prices are tradeoffs, not points. Price, payment model, version numbers and other descriptive values are shown side by side but never scored.
  • Differences move the result. When confirmed answers differ, apps that have the capability gain the weight and apps that don't lose it.

When a report names a pick

A report names one app only when all of these hold:

  • It has the highest net score, with a real gap.
  • The lead rests on a must-have advantage or on at least two factors.
  • No confirmed source shows it missing a must-have.
  • The lead is bigger than everything still unresolved could swing. Each unresolved priority factor counts as twice its weight.

Otherwise the report says there is no clear winner and lists what is still open. Confidence is high when nothing is unresolved and medium when something is.

Limits

  • AI can misread a page. That is why every answer links to its source.
  • Prices and features change. Check the date on a cell before you rely on it.
  • A report reflects the priorities of whoever made it. Use the Research to set your own.
  • Reports are information, not personal advice.

Corrections

Every report has a Report an error button. Point to a page that shows the right information; a source link is required, and each person can send up to five reports a day. A person on our team reviews each one. The outcome and our reply are listed on the report and below, whether we change the report or not. Reporter names and emails are never shown.

Recent corrections

No corrections have been resolved yet.

Contact

Questions about how research works: support@themacapps.com