Reference-grade AI, explained clearly

Our mission

Artificial intelligence changes fast — new models, new architectures, new capabilities every week. Every existing reference we tried failed us in some way: too shallow, too outdated, too self-promotional, or too narrow. So we built AI Terms Guide.

Our mission is simple: explain every AI term, model, tool, and concept clearly enough that anyone — engineer, executive, student, or journalist — can understand and act on it.

We cover the vocabulary of modern AI: 2,000+ terms, 500+ models, 800+ tools, 50+ providers, deep-dive concept tutorials, side-by-side comparisons, profession-specific use cases, and structured learning paths.

What we cover

Everything in AI, organized the way it actually connects.

• Terms. The full vocabulary — from foundational concepts like backpropagation to modern developments like MCP, GRPO, and extended thinking.
• Models. Every major model with real specs, cited benchmarks, current pricing, and honest strengths and weaknesses.
• Tools. Every meaningful AI product reviewed with pros, cons, alternatives, and what it's actually good for.
• Providers. Company profiles for the labs and companies shaping AI — Anthropic, OpenAI, Google DeepMind, Meta, and dozens more.
• Concepts. Long-form tutorials for things that need more than a definition — how transformers work, how RAG works, how agentic AI works.
• Comparisons. Head-to-head decisions with real trade-offs.
• Use cases. AI applications organized by profession.
• Learning paths. Structured lessons from zero to specialist.

Editorial standards

Reference sites live or die on trust. These are the standards we hold every page to.

• Real content only. No invented specs. No fabricated benchmarks. No hallucinated features. If we don't know a number, we say so.
• Cited sources. Every specification, benchmark, and price links to a primary source — a research paper, official documentation, or trusted benchmark aggregator like Artificial Analysis or Chatbot Arena.
• Provider claims vs independent testing distinguished. A vendor's own benchmark is not the same as an independent evaluation. We show both when possible and label which is which.
• Hedged language for judgment calls. "Generally considered" or "in most benchmarks" — never absolutes we can't back up.
• Public corrections. When we're wrong and it's material, we add a dated correction notice to the affected page. Silent edits hide errors and erode trust.
• Balanced coverage. We cover major providers proportionately regardless of who's fashionable. Popularity is not the same as accuracy.

For the full detailed policy, see our Editorial Policy.

How we work

Our workflow has three stages for every page.

1. Research and draft. A writer pulls the primary sources — the paper, the docs, the benchmark data — and drafts the entry.
2. Technical review. An engineer verifies the technical claims, checks the code examples, and catches misstatements before publication.
3. Editorial pass. An editor checks clarity, tone, cross-references, and consistency with the rest of the site.

Only after all three stages does a page go live. Flagship pages (top 50 most-searched terms and models) get additional review by a senior AI engineer.

Who writes and reviews

Two writers lead the editorial: Ahmed R. and Sana K. They rotate authorship across the site's growing library of pages. Both have deep experience translating technical AI content for non-specialist audiences.

Technical review is handled by a rotating panel of AI engineers with hands-on experience shipping production LLM applications — RAG systems, agent loops, fine-tuning pipelines, evaluation harnesses. Reviewer names appear on the pages they review.

Update cadence

AI moves fast. Reference content that isn't updated is worse than no reference at all. We update on a strict cadence:

• Weekly: model pricing changes, new model releases, new tool launches. Every model page has been reviewed within the last 7 days.
• Monthly: benchmark refreshes as new evaluation results come in.
• Quarterly: full audit of every page — clarity, examples, dead links, outdated advice.
• Within 48 hours: new pages for major model or paper releases.

Every page displays its "Last updated" date in the byline. If a page looks stale, ping us on the contact page — we prioritize refreshes based on reader feedback.

How we source information

We treat sources like a research paper does. Not all sources are equal.

Sources we consider authoritative:

• Provider official documentation (Anthropic docs, OpenAI docs, Google AI docs).
• Original research papers (arXiv, published journals with peer review).
• Trusted benchmark aggregators — Chatbot Arena, Artificial Analysis, Hugging Face leaderboards.
• Provider blog posts for company-specific claims (release notes, spec updates).

Sources we avoid citing:

• Random social media posts (unless from official verified accounts).
• Marketing copy without technical substantiation.
• Aggregators that republish company content without adding verification.

Corrections and feedback

We ship corrections within 24 hours of confirming an error. If you spot an inaccuracy — no matter how small — please tell us via the contact page. Material corrections get a dated notice at the bottom of the affected page so readers can see what changed.

How we make money

AI Terms Guide is free to read because it's supported by display advertising and affiliate commissions. Our editorial team is separate from advertising — reviews and recommendations reflect our honest assessment, not commercial relationships. Full details in our Affiliate Disclosure.

Our reference network

AI Terms Guide is one site in a network of specialist AI references, each covering a different slice of the AI landscape:

• AI Error Hub — every AI error decoded and fixed.
• AI Code Toolkit — real code for real AI tasks.
• Skills and Servers — deep coverage of MCP servers and Skills.
• AI Services Cost — real pricing and TCO analysis.
• AI Business Ideas — practical AI-first business ideas.

Each site is independent editorially but shares the same standards for accuracy, sourcing, and updates.

Get in touch

Corrections, partnership inquiries, media requests, or just a note — we read every message. Head to our contact page or reply to any newsletter email.