Explained How uses technology to help research, organize, explain, and visualize complex subjects.

That may include artificial intelligence.

But AI does not determine what we consider true.

Our policy is based on a simple principle:

AI can assist the editorial process. It does not replace evidence, judgment, or editorial responsibility.

Explained How may use AI-assisted tools during research, planning, drafting, editing, analysis, and visual production.

However, important factual claims should ultimately be supported by appropriate sources rather than by the output of an AI system.

The responsibility for content published by Explained How remains with the people identified as responsible for that content.

At present, primary editorial responsibility belongs to Duc Nguyen, founder and editor of Explained How.

Why We Use AI

Explained How covers questions such as:

How does rain form?
How do animals navigate?
How does a jet engine work?
How does the human immune system work?
How does Wi-Fi work?
How does artificial intelligence work?
How does radar work?
How does inflation work?

Researching these subjects often involves large amounts of technical terminology, documentation, interconnected concepts, and potential follow-up questions.

AI-assisted tools can help make parts of that workflow more efficient.

For example, they may help us:

organize research questions;
identify terminology that requires verification;
compare different ways of structuring an explanation;
discover gaps in an outline;
summarize documents for internal review;
organize notes;
improve language and readability;
identify possible misconceptions worth investigating;
develop ideas for diagrams and explanatory visuals.

The purpose of using AI is to improve the editorial workflow.

It is not to avoid doing the underlying research.

AI Is Not a Source of Truth

An AI-generated statement is not evidence.

AI systems can produce information that is:

incorrect;
outdated;
incomplete;
oversimplified;
taken out of context;
confidently fabricated.

They may also invent:

studies;
statistics;
quotes;
experts;
citations;
books;
technical specifications;
URLs.

For that reason, we do not treat an AI response as proof that a factual claim is true.

When an important statement requires verification, we aim to check it against appropriate external evidence.

Depending on the subject, that evidence may include:

government agencies;
scientific institutions;
official technical documentation;
standards organizations;
academic research;
peer-reviewed literature;
engineering documentation;
legislation;
court records;
regulatory agencies;
manufacturer technical documentation.

You can read more about our source hierarchy in:

[How We Research →](/how-we-research/)

How AI May Be Used in Research

AI-assisted tools may help during the early stages of research.

Examples include:

Question Discovery

AI may help generate questions such as:

Which components are necessary for this system to work?

What variables change the outcome?

What assumptions does this explanation depend on?

These questions can help guide research.

They do not determine the answer.

Terminology Discovery

AI may help identify technical terminology related to a subject.

Those terms may then be investigated using authoritative sources.

For example, an AI tool might identify concepts related to:

thermodynamics;
signal processing;
network protocols;
biological pathways.

We would still verify what those concepts mean and how they relate to the mechanism being explained.

Research Organization

AI may help organize:

notes;
source material;
research questions;
competing explanations;
article sections.

This can make complex research easier to manage.

The organization of evidence is not a substitute for verifying the evidence itself.

How AI May Be Used in Writing

AI-assisted tools may contribute to parts of the writing process.

Possible uses include:

preliminary outlines;
alternative explanations;
sentence restructuring;
grammar;
readability;
terminology consistency;
identifying repetitive passages.

However, Explained How does not consider a draft publishable merely because an AI system produced fluent text.

Before publication, the article must still meet our editorial standards for:

accuracy;
sourcing;
causal explanation;
clarity;
originality;
uncertainty;
relevance.

We Do Not Publish Raw AI Output as an Editorial Standard

Our intended workflow is not:

Prompt → Generate → Publish

Instead, it should more closely resemble:

Question

↓

Research

↓

Source verification

↓

Mechanism analysis

↓

AI-assisted organization or drafting when useful

↓

Editorial review

↓

Fact-checking

↓

Revision

↓

Publication

AI may participate in several stages.

It does not eliminate the other stages.

Human Editorial Responsibility

AI does not have editorial responsibility for Explained How.

A language model cannot be held accountable for:

an inaccurate claim;
a misleading explanation;
a fabricated citation;
an inappropriate inference.

The person responsible for publishing the content retains that responsibility.

At present, that responsibility primarily belongs to:

[Duc Nguyen →](/authors/)

Duc oversees Explained How’s research and editorial direction.

As additional real authors, editors, fact-checkers, or reviewers join the publication, their roles will be identified where appropriate.

AI and Fact-Checking

AI may help identify claims that deserve closer scrutiny.

For example, it may flag:

numerical claims;
technical specifications;
dates;
possible contradictions;
terminology;
statements that appear unusually certain.

But AI is not the final fact-checker.

Important claims should be checked against appropriate sources.

If AI output conflicts with an authoritative source, we do not choose the AI answer simply because it sounds clearer.

Our complete verification process is described in:

[Fact-Checking Policy →](/fact-checking-policy/)

We Do Not Use AI to Invent Experience

Explained How does not allow AI-generated writing to create the appearance of experience that did not occur.

Unless real supporting evidence exists, we do not publish claims such as:

“We tested…”

“During our testing…”

“We observed…”

“Our customers found…”

“From our experience…”

AI cannot manufacture first-hand experience for us.

If an article is based on research, it should remain a research-based article.

We Do Not Use AI to Invent Expertise

AI tools may make writing sound authoritative.

That does not make the author an expert.

We do not use AI to fabricate:

qualifications;
professional titles;
academic degrees;
licenses;
employment history;
expert reviewers;
author biographies.

For example, we would not label someone:

“Aerospace Engineer”

or:

“Medical Expert”

unless that description is genuinely accurate and verifiable.

Our current author information is available at:

[Authors →](/authors/)

AI and Originality

Using AI does not remove the requirement to create something useful and original.

Explained How aims to provide value through:

evidence synthesis;
mechanism reconstruction;
causal reasoning;
useful distinctions;
identifying overlooked conditions;
explaining feedback loops;
comparing sources;
clarifying system limits.

We do not want Explained How articles to be generic rewrites of information already available across the web.

A central editorial question is:

What does the reader understand after this article that a generic summary would not have made clear?

AI may help us work.

It should not replace original editorial value.

We Do Not Use AI to Create Content at Scale Solely for Search Traffic

Explained How is capable of covering many subjects.

That does not mean we should automatically generate thousands of pages.

We do not intend to use generative AI simply to create large volumes of:

keyword variations;
near-duplicate articles;
lightly rewritten summaries;
pages produced primarily to capture search queries.

Our editorial mission remains:

Explain how things work.

The number of pages published is secondary to whether an explanation deserves to exist.

Google’s current spam policy similarly defines scaled content abuse around producing many pages primarily to manipulate rankings rather than help users, regardless of whether AI, automation, or humans produced them. Google for Developers

AI and Our “Mechanism-First” Method

Our core editorial model remains:

INPUT → PROCESS → OUTPUT → FEEDBACK → LIMITS

AI may help ask questions around this model.

But each step still needs to make factual and causal sense.

For example:

Input: electricity

↓

Process: compressor and refrigeration cycle

↓

Output: heat moved out of a refrigerator compartment

An AI system might produce a technically plausible description.

Our task is to determine whether that description accurately represents the real mechanism.

This is why source verification remains necessary.

AI-Generated or AI-Assisted Images

Explained How may use AI-assisted images or illustrations.

This is especially useful for explanatory content where a visual can help show:

internal components;
cross-sections;
processes;
invisible forces;
biological mechanisms;
system relationships;
technical concepts.

AI-assisted imagery may be used as:

educational illustration;
conceptual reconstruction;
visual explanation;
supporting artwork.

It should not be treated as documentary evidence.

What AI Images Must Not Pretend to Be

An AI-generated visual should not knowingly be presented as:

an actual photograph of an event;
scientific measurement;
laboratory evidence;
historical photographic evidence;
direct observation;
a real person who does not exist.

If the distinction matters to understanding or credibility, additional labeling or context should be provided.

For example, an illustration showing the internal flow of refrigerant can be useful even if it is partly AI-assisted.

But it should not be described as:

“A photograph of refrigerant moving through this compressor”

when it is only an illustration.

Accuracy Matters in Visuals Too

An image can communicate misinformation even when the accompanying text is correct.

We therefore aim to check whether explanatory visuals correctly represent:

component relationships;
direction of flow;
relative placement;
labels;
causal sequence;
technical concepts.

A beautiful illustration is not useful if it teaches the wrong mechanism.

This is becoming increasingly important as search experiences become more visual and multimodal. Google added dedicated multimodal Search Console reporting in September 2026, covering search experiences involving images and related generative AI features. Google for Developers

AI and Video

Explained How also operates the Explained How YouTube channel.

AI-assisted tools may contribute to video production workflows such as:

research organization;
scripts;
visual concepts;
diagrams;
narration support;
editing;
illustrative imagery.

The same basic rules apply.

AI assistance does not justify presenting fictional evidence as reality.

Official channel:
Explained How on YouTube

Disclosure of AI Use

We do not believe every routine use of software or AI requires a large warning at the top of every article.

However, transparency matters when AI use could reasonably affect how a reader interprets the content.

We may provide additional disclosure when:

AI-generated visuals could be mistaken for documentary imagery;
AI substantially contributed to a particular interactive feature;
automation materially shaped how a page was produced;
the method of creation is relevant to the reader’s trust in the content.

Our general use of AI-assisted tools is disclosed through this policy.

Google’s current people-first guidance similarly recommends providing information about automation or AI when readers could reasonably ask how the content was created, rather than treating disclosure as a blanket ranking tactic. Google for Developers

A Disclosure Does Not Make Low-Quality Content Acceptable

Transparency is not a substitute for quality.

Adding:

“This article was created with AI assistance”

does not make inaccurate, unoriginal, or poorly reviewed content acceptable.

Our editorial standards still apply.

An AI-assisted article must still:

answer the reader’s question;
contain accurate information;
rely on appropriate evidence;
provide useful explanation;
demonstrate editorial effort;
avoid fabricated claims.

High-Stakes Topics

AI-assisted workflows require greater caution for topics involving:

health;
medicine;
personal safety;
law;
finance;
investing;
insurance;
taxes;
elections;
government benefits.

For these topics, AI-generated claims should never substitute for authoritative evidence.

The stronger the potential impact on a reader’s life or well-being, the stronger the verification standard should be.

Where professional review is appropriate, it should be performed by a real qualified person who is identified accurately.

We do not use AI to simulate expert review.

Political and Contested Topics

When AI tools assist research involving politics, government, elections, law, or contested public issues, additional care is necessary.

AI systems can reproduce:

bias;
outdated information;
unsupported claims;
false equivalence;
oversimplification.

We aim to verify relevant claims through primary and authoritative sources and distinguish fact from interpretation.

AI output does not determine Explained How’s political position.

Our purpose is to explain how systems work.

Product and Affiliate Content

AI-assisted tools may help organize product information or technical documentation.

They do not allow us to claim first-hand testing that did not occur.

If Explained How has not personally tested a product, we do not use AI-generated language such as:

“We found the controls easy to use.”

“During our test…”

unless a real test took place.

Affiliate relationships and AI assistance are separate issues.

Neither changes our obligation to describe the basis of our claims accurately.

AI and Sources

We do not intentionally cite:

“According to ChatGPT…”

or another general-purpose AI assistant as primary evidence for a factual mechanism.

The relevant evidence should come from the underlying source.

For example:

Not:

An AI assistant says the protocol works this way.

Prefer:

The relevant RFC describes the protocol this way.

AI may help us discover that an RFC exists.

The RFC is the evidence.

Fabricated Citations

Fabricated citations are unacceptable.

Before relying on a citation suggested by AI, we should verify that:

the source exists;
the title is correct;
the author or organization is correct;
the source actually supports the claim;
the URL points to the intended material.

A citation that exists but does not support the statement is still a sourcing failure.

Quotation Integrity

AI tools can generate quotations that sound plausible but were never spoken or written.

We do not knowingly publish a direct quotation without verifying its source.

When exact wording cannot be confirmed, we should paraphrase the supported information rather than invent quotation marks.

Confidential and Personal Information

We aim to use AI tools responsibly when handling non-public information.

Sensitive information should not be unnecessarily provided to external AI systems.

This includes, where applicable:

passwords;
private account credentials;
confidential correspondence;
unpublished personal information;
sensitive user information.

AI convenience does not override privacy obligations.

Copyright and Attribution

AI tools may generate language or images influenced by large amounts of existing material.

Explained How still aims to respect:

copyright;
attribution;
licensing;
source ownership.

We do not intentionally use AI to disguise plagiarism or reproduce another publisher’s work with superficial wording changes.

When content relies on external evidence, the goal is to synthesize and explain—not simply rewrite.

Corrections Involving AI

If AI assistance contributes to an error, the correction process is the same as for any other editorial error.

We do not consider:

“The AI made the mistake”

an excuse for leaving incorrect information published.

If an error is discovered, we investigate and correct it according to our:

[Corrections Policy →](/corrections-policy/)

How We Evaluate AI-Assisted Content Before Publication

Before publishing AI-assisted material, we ask questions such as:

Purpose

Was this created because it helps readers understand the subject?

Accuracy

Are important factual claims supported?

Originality

Does the article offer meaningful explanation or synthesis beyond generic information?

Mechanism

Does the causal chain actually make sense?

Sources

Have important claims been traced to appropriate evidence?

Experience

Does the content falsely imply testing or first-hand experience?

Expertise

Does it invent credentials or reviewers?

Uncertainty

Are limitations and disagreements represented honestly?

Visual Integrity

Could an AI-assisted image mislead readers about what is real?

Human Responsibility

Has someone taken responsibility for the finished content?

If significant problems remain, the material requires further work before publication.

AI Does Not Decide What We Publish

An AI tool may suggest:

an article idea;
a keyword;
a headline;
a structure;
a related question.

That does not automatically mean Explained How should publish it.

Our editorial decisions should remain based on whether a topic fits the mission:

Help readers understand how something actually works.

We do not intend to publish content merely because an automated tool predicts search traffic.

Google’s current people-first guidance similarly warns against producing content across many subjects primarily to capture search visits or using extensive automation without substantial additional value. Google for Developers

AI, Search Engines, and Explained How

Explained How does not use AI because we believe AI-generated text receives preferential treatment from search engines.

Nor do we avoid AI merely because AI was involved.

What matters is the finished content.

It should be:

useful;
reliable;
original;
accurate;
clearly sourced;
written for people.

Google’s current guidance states that generative AI can be useful for research and content structure, while using it primarily to manipulate rankings violates spam policies. Google also continues to emphasize unique, valuable, non-commodity content rather than special “AEO” or “GEO” tricks. Google for Developers

Our Current AI Responsibility

Explained How is currently founded and primarily edited by Duc Nguyen.

Duc is responsible for determining how AI-assisted tools are incorporated into the editorial workflow and for maintaining the standards described in this policy.

Learn more: