Toolipedia claims to be built from real docs, benchmarks, and search data. This page explains exactly what that means — how tools are chosen, how scores are assigned, where pricing data comes from, and how often everything gets reviewed. Transparency isn't a feature here. It's the baseline.
What Toolipedia is
Toolipedia is a curated, independent directory of AI tools. It is not a product review site, not an affiliate marketplace, and not a lead generation tool for vendors. No tool pays to be listed. No tool pays for a higher score or a better description. The directory exists for one reason: to help people find the right AI tool for what they're actually trying to do.
The current directory covers 100 tools across 23 categories. It is maintained by a single editor and supported by community submissions that go through an editorial review before going live.
How we choose tools
A tool qualifies for inclusion if it meets all four criteria:
- Publicly accessible — available to anyone, not just employees of one company or members of a closed waitlist
- Meaningfully AI-powered — AI must be central to its value, not a minor add-on to an otherwise conventional product
- Verifiable and live — must have a working website, a documented product, and pricing we can confirm
- Distinct — not a clone or white-label of a tool already listed, and not so similar to an existing entry that the difference isn't meaningful
We actively look for tools that are under-discussed relative to their quality — enterprise tools, niche tools, international tools. A large marketing budget is not a criterion for inclusion, and a lack of one is not a reason for exclusion.
How we score tools
Each tool is scored 1–10 across six dimensions. Scores are editorial — they represent the editor's informed judgment based on the inputs below, not a calculated formula. The goal is a honest, comparative signal, not a precise measurement.
Score inputs. For each tool, the editor reviews: the official documentation and feature pages; hands-on use of the free tier or trial; publicly available third-party benchmarks (MMLU, HumanEval, HELM, LMSYS Chatbot Arena, coding leaderboards where applicable); user reviews on G2, Reddit, and Hacker News threads; and the editor's own structured tests against a consistent set of tasks per category.
1 — Writing (1–10)
Measures quality of long-form prose, ability to follow complex style instructions, coherence across extended documents, and practical usefulness for real writing tasks (articles, emails, reports, marketing copy). A score of 10 means best-in-class across all writing types. A score below 5 means the tool produces output that requires significant rewriting before use.
2 — Coding (1–10)
Measures ability to write, debug, explain, and refactor code across multiple languages. Informed primarily by HumanEval, SWE-bench, LiveCodeBench, and structured tests across Python, JavaScript, and SQL at varying difficulty. A score of 10 means the tool consistently solves hard coding problems correctly. A score below 5 means errors are frequent enough that the output cannot be trusted without careful review.
3 — Research (1–10)
Measures factual accuracy, citation reliability, ability to synthesize across sources, and usefulness for real research tasks. Tools that hallucinate frequently, refuse to cite sources, or cannot handle long documents are scored lower. Tools with live web access, strong grounding mechanisms, or purpose-built research pipelines score higher.
4 — Creativity (1–10)
Measures range of creative output, quality of image or video generation, originality of ideas, and ability to produce unexpected, non-generic results. For writing tools, this tests poetic quality, narrative creativity, and tonal range. For image tools, this tests aesthetic distinctiveness against a consistent set of prompts. A 10 means the output is genuinely surprising or impressive. A low score means the output is competent but forgettable.
5 — Automation (1–10)
Measures the tool's ability to execute multi-step tasks without supervision, connect to external services, trigger workflows, handle errors, and run reliably at scale. Tools that require a human to prompt every step score low. Tools with agents, pipelines, API-first design, or deep integration ecosystems score high.
6 — Multimodal (1–10)
Measures the breadth and quality of non-text capabilities: image understanding, image generation, audio, video, code execution in context, PDF reading, and so on. A tool that only handles text scores 1–3. A tool with strong vision plus one other modality scores 5–7. A tool with strong performance across vision, audio, and code in a single context scores 8–10.
Avoiding score inflation. Scores are calibrated against each other, not assigned in isolation. Before publishing a new tool, its scores are reviewed against the full directory to check for drift. The aim is that a 7 in Writing means roughly the same thing across all tools in the directory. Scores are revised when there is material evidence of capability change — a major model update, a new benchmark result, or a consistent pattern in user reports.
How we track pricing and plans
Pricing data is sourced exclusively from the tool's official pricing page. Blog posts, reseller sites, and press releases are not used as primary sources. For tools without a public pricing page (enterprise-only), we note this explicitly and provide estimates from public filings, press coverage, or user-reported figures with a clear "estimated" label.
- Annual vs monthly billing — where a tool offers both, we show the monthly cost of the annual plan (the more common purchase path) and note that monthly billing costs more. We do not hide the annual commitment requirement.
- Free tiers — we distinguish between a genuine free tier (no credit card, indefinite use) and a free trial (time-limited or credit-limited). The free badge on a card means the tool has a usable free tier, not just a trial.
- Enterprise pricing — where the vendor does not publish pricing, the card notes "Custom pricing — contact sales" rather than showing a number we cannot verify.
- Last verified — pricing changes without notice. Every tool card shows the month and year the pricing was last checked. If more than three months have passed since the last check, we flag the entry for review.
How often we update
The directory runs on a rolling review schedule:
- Monthly — pricing and free tier details for the top 10 most-visited tools
- Quarterly — full pricing audit across all 100 tools; score recalibration for any tool that has shipped a major model or feature update
- As needed — immediate update when a tool announces a significant change (new model release, pricing restructure, feature removal)
- Community-triggered — any verified correction submitted through the community queue is reviewed within two weeks
The footer of every page shows the current data snapshot date. Individual tool cards will show a "Last verified" month in a future update.
How you can suggest corrections
If you find an error — wrong pricing, an outdated feature, a missing integration, an inaccurate score — use the Submit tab in this Community section. Select the existing tool name, describe what's wrong and what the correct information is, and include a source link if you have one. The editor reviews all submissions before applying any change.
What helps: a direct link to the official pricing or documentation page, a benchmark result with a citation, or a specific feature that the current description misses or misrepresents.
What doesn't help: "the score feels low" without evidence, comparisons based on a single personal use case, or requests to raise a score based on information you cannot publicly source.
Score tooltip copy
The following text is used in score tooltips and sidebar blocks throughout the directory:
Short version (tooltip / sidebar):
Scores are editorial — based on the editor's structured tests, publicly available benchmarks, and hands-on use of each tool. They reflect relative performance across the directory, not absolute measurements. A 7 in Coding means this tool is better at code than most in the directory. Pricing is verified from official pages and dated by month. Neither scores nor pricing are guaranteed current — see the full Methodology for update frequency and how to report corrections.
Microcopy (card footer):
Scores: editorial, from tests & benchmarks · See Methodology
Methodology last updated: March 2026. This document will be revised when scoring criteria or update processes change materially.