An Independent Reviewer Spills the Beans: How I Personally Test AI Tools Before Writing a Review

AI tool reviews are everywhere, but how many are actually written after hands-on use? AIX Radar founder M. Hasnain has made his review process public — no relying on vendor press releases, no selling rankings, every step run through himself. This approach has reference value for both content creators and tool selectors.

Why the Review Process Itself Deserves Attention

AI tools iterate fast, new products emerge endlessly, and the biggest problem users face isn't "is there a tool" but "is this tool actually good to use." A large number of review articles are essentially rearrangements of official introduction pages — after reading them, readers still have no idea what the actual experience is like.

M. Hasnain's approach returns to a simple logic: a reviewer should first and foremost be a user. He makes it clear that every review starts with him registering an account, running real tasks, taking real screenshots, and only then drawing conclusions. The process sounds simple, but its execution cost is far higher than rewriting a press release.

Breaking Down the Four-Step Review Method

Step One: Register as an Ordinary User

No media accounts, no special channels. This step seems basic, but in practice it filters out a large number of "looks usable" misjudgments. The permissions, quotas, and even interfaces vendors give to the media may be completely different from what ordinary users see. Registering as an ordinary user means the limitations, paywalls, and onboarding flows the reviewer experiences match those of real users.

Step Two: Test with Real Workloads

This is the most critical step in the entire process. Hasnain emphasizes using "real work, not demo prompts" — not running through official example prompts to see if the output looks pretty, but throwing the tool into his own daily workflow. The problem with demo prompts is that they're usually carefully selected to maximize the display of a tool's strengths while avoiding edge cases and failure scenarios.

Real tasks test more dimensions: does long-text processing lose context? Does quality degrade after multiple rounds of conversation? Are export formats usable? These only get exposed when you actually use the tool to get things done.

What Is a Demo Prompt?

A demo prompt refers to example prompts vendors display on their official websites, at launch events, or in promotional materials, typically polished repeatedly and specifically designed to showcase a tool's best performance on a certain type of task. Their characteristics are clear tasks, clean context, and outputs that easily stand out. But in real use, users face vague requirements, messy inputs, multiple rounds of revision, and repeated follow-up questions — scenarios that demo prompts precisely do not cover. Therefore, testing only with demo prompts amounts to validating a tool solely under the optimal conditions set by the vendor, which easily overestimates its actual capabilities. A more reliable approach in reviews is to introduce long documents, colloquial instructions containing typos, and tasks requiring multiple back-and-forth exchanges, then observe whether the tool loses context, whether outputs are stable, and whether formats are usable.

Step Three: Compare Pricing with Actual Delivery

Many reviews only list price tables without asking "what features does this price point actually correspond to." Hasnain's approach is to compare what you actually get at each tier against the price — what the free version cuts, what the Pro version adds, whether the team version is worth the difference. For usage-based tools, he also estimates the real monthly cost under typical usage scenarios.

This step directly addresses users' most practical decision-making question: what am I getting back for the money I spend?

Usage-Based Billing and Real Monthly Costs

AI tool billing generally falls into two categories: subscription-based (fixed monthly/yearly fees) and usage-based (metered by tokens, number of calls, generation time, etc.). Subscriptions are easy to compare, while usage-based billing hides cost differences caused by usage habits — with the same tool, a light user might spend only a few dollars a month, while a heavy user might spend hundreds. Estimating real monthly costs requires first tallying your typical usage: how many calls per day, roughly how many characters per input and output, whether images or long audio are involved, then multiplying by unit price. Many tools provide usage dashboards or historical bills that can be referenced directly. For ordinary users, if usage-based billing is hard to predict, it's better to first run real tasks for a week using the free quota, record consumption, and then decide whether to upgrade — this gets closer to actual spending than just looking at a price list.

Step Four: Publish Conclusions and Verified Promo Codes

The endpoint of a review is a clear judgment: what's good, what's bad, and whether it's worth paying for. Hasnain also mentions including promo codes he has verified himself. This detail illustrates the trust relationship between the reviewer and readers — if a promo code is posted without verification, it damages the credibility of the review itself.

The Business Model Dilemma of Independent Reviews

Hasnain draws a clear line in his article: no paid rankings, no rewriting press releases. This precisely points out a structural problem in the AI tool review space.

Reviews themselves have costs — time, subscription fees, opportunity cost. If completely non-commercialized, sustained output becomes unsustainable; if reliant on vendor sponsorship, independence comes into question. Hasnain's chosen path is to operate an independent site, generating revenue through verified promo codes and similar means while maintaining editorial independence. Whether this model can scale depends on whether readers are willing to pay for trustworthy reviews.

For readers, judging whether a review is trustworthy can rely on several signals: whether it shows screenshots of the reviewer registering and using the tool, whether it points out the tool's shortcomings, and whether it distinguishes the actual differences between pricing tiers. If a review is all positives, all official wording, and has no trace of actual use, then it most likely hasn't undergone real testing.

The Boundary Between Affiliate Marketing and Editorial Independence

Generating revenue through verified promo codes or affiliate links falls under affiliate marketing in the content industry. This model itself doesn't necessarily harm independence — the key is whether revenue is tied to review conclusions. If a reviewer tests all tools equally and only attaches promo codes to products genuinely worth recommending, then revenue comes from readers' purchasing behavior rather than targeted vendor sponsorship. The risk is that when a particular vendor's commission rate is noticeably higher, the reviewer may unconsciously lean toward recommending it more. Therefore, the transparent approach is to disclose revenue sources, explain whether affiliate links are used, and ensure negative conclusions are also fully presented. For readers, seeing a promo code doesn't require immediate dismissal, but you can check whether the review also points out shortcomings and distinguishes tiers, to judge whether the recommendation comes from genuine experience or commission-driven motives.

Implications for AI Tool Selection

This process isn't just for reviewers — ordinary users can borrow the same approach when choosing tools.

First, try tools with your own real tasks rather than looking at official demos. Most AI tools have free quotas or trial periods — running through the type of task you do most often is more useful than reading ten reviews. Then compare pricing and calculate clearly how much you'd spend under your actual usage — don't be misled by "the free version works." The gap between free and paid versions is often what determines the experience.

Finally, maintain a bit of caution toward any review: check whether it shows real usage, whether it mentions shortcomings, and whether it distinguishes tiers. Reviews that do all three are much more trustworthy.

One-line Summary

AIX Radar founder M. Hasnain has made public a four-step independent review method — no press releases, no selling rankings — offering a reusable validation approach for AI tool selection.

Paragraph Summary

The article introduces AIX Radar founder M. Hasnain's AI tool review process, the core of which is that a reviewer must first be a real user. The process has four steps: register as an ordinary user to avoid misjudgments caused by media privileges; test with real workloads rather than official demo prompts to expose edge-case problems such as long-text context loss, quality degradation in multi-turn conversations, and unusable export formats; compare pricing with actual delivery to figure out what you really get at each tier and the monthly cost under typical scenarios; and finally publish clear conclusions with verified promo codes. The article also points out that independent reviews face business model challenges — complete non-commercialization is unsustainable, while vendor sponsorship raises independence concerns, and verified promo codes are a compromise path. Ordinary users can borrow the same approach when choosing tools, and should be wary of reviews with no trace of actual use.