- Prices are read on the vendor’s own pricing page, dated, in the vendor’s currency.
- Every review says who a plan is worth it for, who should skip it, and which alternatives to consider.
- Comparison scores are editorial judgments with a written note for each line, not averages of user reviews.
- Prompt pages are built around a real job, with a description and a tip for every prompt.
- Sponsors, ads and vendor requests never change a verdict, a score or a pick.
1. “Is it worth it” reviews
AI Academy has 31 reviews of paid AI plans. We review the plans readers are deciding whether to pay for, such as ChatGPT Plus, Claude Pro, Cursor or Midjourney, and every review follows the same structure:
- The answer first: a short verdict with the price, before any background.
- What the money buys: the paid tier against the free tier and the vendor’s other plans, in numbers (message caps, credits, limits).
- A comparison table of the plans, read on the vendor’s pricing page.
- Pros and cons, including the real weaknesses.
- Worth it for / skip it if: the people the plan fits, and the people who should keep their money.
- Alternatives by name, with what each one does better.
- An FAQ and links to the vendor pages the facts were checked on.
Reviews do not show a star score. Each review carries a 1 to 5 rating in its structured data for search engines, but the verdict readers see is written out, because “worth it” depends on how you use the tool.
2. Head-to-head comparisons
AI Academy has 14 tool-versus-tool comparisons. Each one shows:
| Element | What it is |
|---|---|
| Pricing tiers | Each tool’s plans side by side, from the vendors’ pricing pages. |
| Best for / weakness | One line each on where the tool shines and where it falls short. |
| Capability scorecard | Capabilities scored 0 to 10 for each tool, each line with a note explaining the score. |
| Use cases | Common jobs with a winner for each, or a tie when neither tool is clearly better. |
| Overall rating | A rating out of 5 for each tool. |
| Verdict | Which tool to pick, and for whom. |
Scores and ratings are editorial judgments made from the vendor documentation and pricing we checked. They are not averages of user reviews and are not produced by a formula, which is why every scorecard line carries a note you can argue with.
3. Prices
- Read on the vendor’s own pricing page for the page being written, never recalled from an older article.
- Dated, so you know when the number was true.
- In the currency the vendor lists. We do not convert.
- No third-party estimates. If a vendor publishes no price, the page says to check current pricing.
AI plans change often. When a vendor changes a price or a limit, the vendor’s page wins and ours gets fixed: see corrections.
4. How we pick and write prompts
The prompt library holds 45,689 prompts on 1,752 pages. A page is built around a job or a role people actually search for (a resume, a cold email, a YouTube script), for a specific assistant such as ChatGPT or Claude. Pages written under the current rules follow a fixed format:
- 30 prompts in six use-case groups of five.
- A two-sentence description and a practical tip for every prompt.
- Bracketed placeholders, such as [YOUR NOTES] or [YOUR PHOTO], where you add your own details.
- A guardrail line where the prompt works from your material, such as “Use only what I paste. Do not invent a number, quote, date, ticker, or source.” The assistant then flags gaps instead of filling them.
- Honest limits on sensitive topics: no prompts designed to get around an assistant’s safety policies.
- A short FAQ and links to related prompt pages and role pages.
Before a prompt page ships, automated checks confirm the prompt count matches the prompts on the page and that every internal link works and points to another page.
5. Tutorials
A tutorial is judged on whether a beginner can follow it to the promised result. Our pre-publication checklist walks the steps in order, checks every screenshot against its step, tests that each prompt produces what the step claims, verifies product names and features against the vendor’s official documentation, and requires the last step to show the finished result. See the editorial policy for the full list of blockers.
6. What never influences a verdict
- Sponsored placements. They are labelled and sold separately. They do not change a review, a score, a comparison winner or a prompt pick.
- Ad slots. An ad shown on a page has no link to what the page says.
- Newsletter sponsorships. A company that sponsors Techpresso or another Dupple newsletter gets no different treatment on AI Academy.
- Vendor requests. A vendor can correct a fact with a link to its own page. It cannot buy or negotiate a verdict.