Substantiated.ai exists because most AI tool coverage is either vendor marketing dressed up as editorial, or surface-level listicles that tell you features exist without telling you whether they work. This page explains exactly how we evaluate tools and how our verdicts are formed.
We review tools that business teams are actively evaluating or already paying for. The selection criteria are simple: is this something a real team is spending real money on, or considering spending money on? If the answer is yes and we haven't covered it, it's on the list.
We don't accept review requests from vendors. We don't take payment to prioritise a tool, expedite a review, or cover a product launch. The review queue is editorially determined, not commercially determined.
We focus on the use cases that most business teams actually encounter - not edge cases or enterprise-only configurations. If a feature only works at a $50k+ contract tier, we say so rather than reviewing it as if it's accessible to every buyer.
Every review ends with one of two verdicts: Recommended or Conditional. We don't use star ratings, numerical scores, or "best for" superlatives because those formats reward surface-level evaluation and obscure the use-case specificity that actually matters.
The tool does what it claims, performs well against the use cases we tested, and the pricing is defensible relative to the value delivered. We'd tell a colleague to try it without caveats - though we always note who it's not right for.
The tool has genuine value but only in specific circumstances - usually a particular scale, use case, or budget threshold. We explain the condition precisely. "Conditional" is not a soft no; it's a specific yes with a defined qualifier.
Every verdict includes a "not ideal for" section. A recommendation with no caveats is an ad. We describe the scenarios where a tool fails, underperforms, or where a cheaper or simpler alternative would serve the reader better. This is non-negotiable regardless of how good the tool is overall.
We also try to be specific about the price point at which a verdict changes. A tool that's great value at $49/month may be hard to justify at $249/month. We say so rather than leaving the reader to figure it out.
Our evaluations are built on independent research. We draw on verified user feedback from platforms like G2 and Capterra, published pricing and feature documentation, credible third-party benchmarks and comparisons, and real-world reporting from teams that have deployed these tools at scale.
This means we're synthesising signal from a broader base than any single test environment could produce. Hundreds of verified reviews from actual business users tell you more about day-to-day reliability than a single reviewer's session. We weight that evidence accordingly and look for patterns across sources rather than treating any one data point as definitive.
What we don't do: We don't rely on vendor documentation, demo videos, or marketing materials as evidence. We don't treat a vendor's published benchmarks as fact — where they're cited, we assess them critically against independent sources. If a claim can't be substantiated from outside the vendor, we don't present it as our finding.
When we include numbers in a review — accuracy figures, time savings, performance comparisons — we describe exactly where they came from. If a figure originates with the vendor and we couldn't independently verify it, we say so explicitly.
We do not accept payment from vendors to cover, prioritise, or positively frame any tool. We do not send reviews to vendors for preview or approval before publication. What you read is what we found - not a version that's been negotiated with a PR team.
We do not accept free access to paid tiers in exchange for coverage. If a tool has a free plan, we evaluate it on the free plan before deciding whether to test a paid tier - which we access at our own cost.
We will occasionally be contacted by vendors after a review is published. We update reviews when products substantively change - not in response to vendor feedback about our conclusions. If a tool we reviewed negatively ships a significant update that addresses our criticisms, we'll re-evaluate and update the review accordingly.
No vendor pays to be reviewed, prioritised, or framed positively. The review queue is editorial.
Reviews are never shared with vendors before publication. We don't negotiate conclusions.
We don't accept free paid tiers in exchange for coverage. We evaluate tools at our own cost.
Reviews are updated when products change substantively - not in response to vendor complaints.
Some reviews contain affiliate links. If you click one and purchase a paid plan, we may receive a commission from the vendor at no additional cost to you. Affiliate relationships do not influence our verdicts. Tools we don't recommend still carry affiliate links where they exist - because the link is a navigation convenience, not an endorsement signal.
We will never recommend a tool we don't believe is genuinely useful in order to earn a commission. The long-term value of this site depends entirely on readers trusting that our verdicts are honest. A single compromised recommendation would undermine that. The incentive structure runs in the opposite direction: we have more to lose from a dishonest positive review than we have to gain from the affiliate revenue it might generate.
Every page that contains an affiliate link discloses this in the disclaimer section at the bottom of the review. The disclosure is specific - it names the vendor and describes the relationship plainly. We don't bury it in a footnote or use language designed to obscure what it means.
If a tool we cover doesn't have an affiliate programme, the review looks identical to one that does. There is no financial incentive to prioritise tools with affiliate programmes over those without.
Every review follows the same structure: a summary card with the at-a-glance verdict, a description of what we tested and how, the specific findings broken down by feature area, a direct comparison where relevant, a pros and cons section, and a final verdict with explicit reasoning.
The most important section for most readers is "Who it's right for - and who it isn't." This is where we try to be most specific about the conditions under which our verdict applies. A recommended tool for a 20-person marketing team may be the wrong choice for a solo operator - and we try to make that clear rather than leaving the reader to infer it.
If your specific use case isn't covered in the review, treat the verdict as provisional until you've tested it against your own workflow. We describe what we evaluated. If your workflow is meaningfully different, our conclusions may not transfer directly.
A note on recency: AI tools move fast. A verdict that was accurate at time of publication may be outdated if the product has shipped significant updates. We note the evaluation period at the top of each review and update when we're aware of material changes. If you're reading a review that's more than six months old, check the vendor's changelog before relying on specific feature assessments.
No schedule, no filler. We send when something is worth reading.