Research files: how to source content where accuracy matters

A research file is a document built before any writing starts that lists every fact an article may use, each with its source URL and the date someone checked it. Writers state only what is in the file. Anything nobody could verify goes on a separate list and stays out of the article.

It is a dependable way to keep health, legal and financial content accurate, and it works at every funnel stage: TOFU (top of funnel) explainers, MOFU (middle of funnel) comparisons and cost pages, and BOFU (bottom of funnel) service pages. This guide explains what Google means by YMYL, what goes in a research file, which sources count, and how the file connects to review and to the citations readers see.

What YMYL means, in Google's words

The current Search Quality Rater Guidelines are dated September 11, 2025. Section 2.3 says: "Some topics have a high risk of harm because content about these topics could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society. We call these topics 'Your Money or Your Life' or YMYL."

The guidelines group the harm into four types: YMYL Health or Safety; YMYL Financial Security; YMYL Government, Civics and Society; and YMYL Other. They also say "Many or most topics are not YMYL." A guide to choosing paint colors is not YMYL. A guide to ADHD medication, tax payment plans or a legal deadline is.

Two cautions on how this works. Rater ratings don't move rankings directly; Google says "Rater data is not used directly in our ranking algorithms." And E-E-A-T (experience, expertise, authoritativeness and trustworthiness) "itself isn't a specific ranking factor." What Google does say on its people-first content page is that its systems give "even more weight to content that aligns with strong E-E-A-T" for these topics, and that YMYL content "must be highly accurate and consistent with established expert consensus." Our explainer on YMYL and E-E-A-T covers the background.

What a research file is

A research file is a sourced list of facts for one topic area, built before writing. Every line has four required parts and one optional one:

The file also has two lists at the end. The first is "could not verify": claims someone wanted to use but could not confirm on a first-party page. The second is "do not claim": common myths and outdated statements that look true but aren't. Both stay out of the article.

The rule that makes the file work is simple. Writers may state only what is in the file. If a writer needs a fact that isn't there, it gets researched and added first, or it doesn't go in.

Which sources count

A first-party source is the organization that owns the fact. For most YMYL topics that means:

• Regulators and agencies: the agency's own rule text, guidance or press release.
• Statutes and regulations: the official text, or the government's own online edition, such as the Electronic Code of Federal Regulations (eCFR) for federal rules.
• Standards and professional bodies: the body's own published guideline or rule.
• Vendors: the product's own documentation, help center, pricing or policy page.
• Insurers and payers: the plan's own coverage policy.
• The business itself: for its own services, prices and credentials, confirmed in writing.

Secondary sources, such as news articles, blogs and other agencies' summaries, are useful for finding the primary source. They aren't the source. If a fact appears only in secondary sources, it goes on the could-not-verify list. That rule keeps out most of the invented statistics that circulate between websites.

Record the page's own "last updated" date where it has one, as well as the date you checked. Some sources can't be fetched directly. If a page blocks access, note that, and have someone confirm it from another network before publishing.

The client-confirmed facts file

Facts about the business are kept in a separate file: services offered, prices, locations, hours, insurance accepted, credentials and licenses, who provides each service. Each line records who confirmed it and when.

These facts change more often than regulations do, and getting them wrong causes real problems: a page promising a service the business dropped, or a credential the practitioner doesn't hold. Google's people-first page now calls fabricating creator profiles, for example with "made-up names, or false credentials," "a form of deception." That line is about authors, but the same care applies to credentials claimed for a business. A confirmed-facts file is how you avoid publishing a false one by accident.

When a fact isn't confirmed, write around it. Say "ask about insurance when you call" rather than listing plans nobody confirmed.

Dated facts and recheck dates

Some facts in a research file have a known expiry. Mark those with a recheck date. Two examples from eSEOspace's own guides: on an adult ADHD guide, a telehealth prescribing flexibility set to expire December 31, 2026 was noted for recheck; on a window-replacement guide, a state law taking effect January 1, 2027 was noted the same way.

The recheck date belongs in the file, and the date-sensitive wording belongs in the article ("as of October 2026"). The full routine for this is in keeping a content guide current.

How eSEOspace uses research files

In eSEOspace's process, each topic area gets a research file before any article in it is written. Every line has a first-party source URL and a check date, and covers whatever the topic needs: regulations, insurance rules, standards, vendor documentation and the client's confirmed facts. On one guide for a New York City psychiatry practice, each article averaged about 10 sources.

That is a description of method, not a claim about results. The file is there to make the articles accurate and checkable, which is the job no matter who writes them.

The independent check

A research file prevents many errors. A second pass catches what slips through. Someone who didn't write the article goes through it line by line and checks each claim against the file, and the file against the live source where anything looks off.

1. Highlight every factual claim in the draft: numbers, names, dates, rules, product features.
2. Find each one in the research file. Anything missing gets flagged.
3. Check quoted wording matches the source exactly.
4. Check nothing from the could-not-verify or do-not-claim lists slipped in.
5. Check hedges survived editing. Trimming often turns "may" into "will."
6. Send the checked draft to the subject expert for review.

On medical, legal and financial pages, expert review comes after this check, and the pages stay out of Google's index until it's done. That process is covered in expert review before indexing. The editor's wider role is discussed in the role of human editors.

Research files and AI drafting

If you use AI tools in drafting, the research file matters more, not less. Google's guidance on generative AI content says: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." It adds that the same check applies to metadata: title elements, meta descriptions, structured data and image alt text.

A file of verified facts gives the drafting step something to stay inside, and the independent check confirms it did. Google's position on AI itself is neutral: "Appropriate use of AI or automation is not against our guidelines." The problem Google names is content made primarily to manipulate rankings, such as many pages generated without adding value. Producing many thin pages quickly is the subject of big content guides and the scaled content policy.

Citations readers can see

The research file is internal, but the article should show its sources too. Link the first-party source for each important fact, close to the claim. Google's link guidance notes that external links "can help establish trustworthiness (for example, citing your sources)." Link to the specific page, not a home page, and keep the count sensible; a few strong sources beat a long list.

Visible citations also help readers check you. That is the point of YMYL care: a reader making a health or money decision should be able to see where each fact came from.

Conclusion

A research file turns "be accurate" into a working rule: every fact sourced to a first-party page and dated, unverifiable claims kept out, business facts confirmed in writing, and a second person checking the draft before an expert reviews it. It takes time up front and saves far more later. The TOFU, MOFU, BOFU content strategy pillar shows where research sits in the full method. If you'd like help building research-backed content for a YMYL topic, see our content marketing services. More guides are in our GEO and AEO Guide.

Frequently asked questions