How We Test and Score AI Tools
Every GuglyTech review follows the same process: real accounts, real tasks, claims checked against official sources, and a score built on a fixed rubric. Here is exactly how it works.
Our testing process, step by step
No tool gets a score until it has gone through all six steps below.
Sign up like a regular user
We create a normal account on the public website. We use the free tier first, then a paid plan when the paid features matter to the verdict. No vendor demo accounts, no special access.
Define who the tool is for
We decide the core use case first — a blogger drafting posts, a marketer making ad creatives, a student checking text. The tool is judged on that job, not on its marketing page.
Run a standard task set
Every tool in a category gets the same prompts and projects. That keeps comparisons fair when we put tools side by side.
Verify every claim
Pricing, limits, model versions, and privacy terms are checked against the tool’s official pricing page, docs, and terms of service. If a number isn’t published, we say so instead of guessing.
Score with a fixed rubric
Results go into the same weighted scorecard for every tool, so an 8.2 means the same thing across reviews.
Re-test when things change
When pricing, plans, or core features change, we re-run the relevant tests, update the review, and change the “last updated” date.
What we test in each category
These are the standard tasks every tool in a category goes through.
| Category | Standard tasks | What we look for |
|---|---|---|
| AI chatbots & assistants | Research question, long-document summary, multi-step reasoning, coding snippet, follow-up memory | Accuracy, hallucinations, context handling, speed |
| AI writing tools | Blog intro, 1,000-word draft from a brief, rewrite in a set tone, product description | Originality, factual errors, edit time needed, tone control |
| AI image & video | Same prompt set: product shot, portrait, text in image, short clip from a script | Prompt adherence, artifacts, render time, commercial-use terms |
| AI detectors | Samples of fully human text, fully AI text, and AI text edited by a human | False positives on human writing, consistency across runs |
| AI character & chat platforms | Setup, model/API connection, long conversation, content controls, account settings | Stability, setup difficulty, safety and privacy controls |
How the Editor’s Score works
Every review ends with an Editor’s Score out of 10. It is a weighted average of six criteria.
| Criterion | Weight | What it measures |
|---|---|---|
| Output quality | 30% | How good and accurate the results are on our standard tasks |
| Value for money | 20% | What you get on the free tier and each paid plan versus close competitors |
| Ease of use | 15% | Onboarding, interface, and how fast a new user gets a useful result |
| Features | 15% | Depth of tools, integrations, exports, and customization |
| Privacy & safety | 10% | Data use policy, training opt-outs, content controls, account security |
| Reliability & support | 10% | Uptime during testing, bugs, error handling, help docs and support response |
What we always do
- Publish the cons. Every review lists what underwhelmed us.
- Link to sources. Pricing and policy claims link to the official page.
- Show dates. Each review shows when it was last checked.
- Fix errors fast. Corrections are made and noted in the article.
What we never do
- Sell scores. No company can pay for a review, a higher score, or a better ranking.
- Let vendors edit. Tool makers never see or approve a review before it goes live.
- Review from spec sheets. No hands-on test, no score.
- Invent numbers. Unpublished data is marked as unpublished.
How GuglyTech makes money
GuglyTech is free to read and is supported by display advertising. We may also publish sponsored content. Any sponsored article is clearly labeled as sponsored at the top of the page, and it is kept separate from our reviews and scores. Advertisers and sponsors have no say in what we test, what we write, or how we score.
For full details, see our Disclaimer.
Limits of our testing
AI tools change fast, and their output varies from one run to the next. Our results reflect the plan, model version, and date we tested, from a US account. Your results may differ with different prompts, plans, or regions. Pricing is in US dollars and can change without notice, so always confirm on the official site before you buy.
Spot an error or want a tool reviewed?
If a price changed, a feature moved, or we got something wrong, tell us. We check every message.
Contact us →