How Choice researches companies

How companies are selected

Choice began with a small group of well-known companies added by hand.

After you explicitly allow domain checks, the extension asks Choice for an existing company record when you visit a website. It sends the bare domain and nothing else, such as target.com — never the full web address, the page you are on or anything you type. There is no account and no login. If Choice has no record, it adds only that bare domain and an aggregate request count to the research queue; it does not store who asked or a browsing history. See what Choice never sees.

How sources are found

Choice searches public records for clear examples of what a company said or did. It looks at news, the company’s own website and original records linked by other sources.

The same story may appear on several websites. Choice groups copied reports together so one story is not counted several times.

Which people Choice researches

Choice researches people who can make or materially influence a company’s direction—not everyone who works there. That includes current chief executives and executive chairs, current board members, controlling shareholders, and investors with documented rights such as appointing directors or approving major company decisions. A founder’s historical connection is shown separately from any current authority.

A job title, investment or shareholding is not enough by itself. Ordinary employees, passive investors and non-controlling shareholders are excluded unless a source documents a specific governance or decision-making role. Choice keeps the exact title used by the source and explains why each included person belongs in the company record.

Every included person receives the same bounded personal-action research whether the available evidence is favorable, unfavorable, mixed or empty. Company actions are never assigned to a person merely because they lead the company, and personal actions are never presented as company actions.

Scope is not interchangeable. Catalog contains definitions: company checks and person checks. Organization-only checks recorded as not applicable for a person were not researched as person actions. Each category and definition’s scope appears in the public catalog.

How research becomes a finding

1. Match the right company

   Choice checks the website, brand and owner so a story is not given to the wrong company.

2. Find a clear claim

   Choice looks for something the company said or did—not an opinion or a vague impression.

3. Save the source

   Choice saves the exact words, link, publisher and date so you can check the evidence yourself.

4. Sort it by issue

   Choice connects the finding to an issue such as privacy, workers, climate or political spending.

5. Check the match

   Choice checks that the source supports the finding and names the right company.

6. Group the same story

   If several articles describe the same event, Choice groups them together. Copies do not count as extra proof.

How Choice uses automation

Choice uses automated tools, including AI, to find public sources, read them, sort them and connect them to the right company and issue.

Before a company page appears in search results, automatic checks make sure it has enough findings from enough different sources. AI cannot publish a weak page just because it sounds confident.

Automation can make mistakes. It can misunderstand a source, pick the wrong date, repeat the same story or connect something to the wrong company. That is why every finding links to its source and Choice has a public feedback form.

What a company page needs

A company page does not appear in search results until it has all of these:

• Four sourced findingsA page with less than that under it names a company and says almost nothing, so it stays out of search until the research fills in. Until then it is published as an early record and says so at the top.
• A checked source for every findingChoice must have opened the page and matched the finding to exact supporting words on it. Search headlines and snippets do not count as evidence.
• The required level of proofEach issue has its own evidence rule. Where independent or multiple-publisher confirmation is required, company-controlled sources or copies of one report cannot satisfy it.
• A verified identity, logo and sectorThe page must belong to the right company and have the information needed to identify and compare it.

If an initial result is thin or concentrated in too few issues, Choice runs one bounded round of more targeted research. Until the complete publication rules pass, the page stays out of public search results. Passing means the current research run has enough verified evidence to publish; it does not mean every fact about the company has been found.

Balance across stances is an aim, not a gate. Choice would rather show findings against a company, findings for you to decide on and findings in its favour, in comparable numbers. Public records do not divide evenly: regulators publish enforcement in detail and good behaviour hardly at all, and coverage is far thinner for companies that are private, foreign, or in industries American regulators barely touch. Requiring an even split would hide almost every page and would tell you more about which registers exist than about the company. So the split is not enforced, and a page shows how many kinds of issue its record covers, how much of it is recent and whether anything is in the company’s favour, so you can see the shape of what was found rather than only its total.

You choose what matters

Choice does not decide whether a company is “good” or “bad.” Company pages show sourced actions by topic. You choose the issues that matter to you, and the extension shows how each company lines up with those choices.

Your choices only change which findings matter to you. They never change the company record. Companies cannot pay to be included, removed, moved higher or shown more favorably.

Those choices stay in your browser. Choice never receives them, so Choice cannot see which issues you picked.

What Choice never sees

Choice works without knowing who you are. There is no account, no login and no profile, and the values you choose never reach our servers.

The extension makes no request and reads no site domain until you explicitly allow domain checks in its first-run screen. Declining keeps every network feature off. You can withdraw that choice later by disabling or uninstalling the extension, which stops the requests and takes its stored data with it.

• Only the bare domain. The extension sends the domain by itself — target.com — and nothing more. Never the full web address, the page path, what you searched for, what is on the page or anything you type into it. The server rejects any request carrying more than a bare domain.
• No account, no login, no identifier. A lookup carries no account, no sign-in, no device ID and no tracking cookie. There is nothing in it that points back to you.
• Lookups are never logged. The domain is not written to a log, a history or a per-person record. Choice stores company research, not who looked at what — so there is no browsing history here to leak, sell or hand over.
• Your values stay on your device. The issues you choose to be warned about are saved in your own browser and never leave your device. Choice cannot see what you selected, because Choice never receives it.
• The matching happens on your device. Your browser compares the company’s evidence against your issues locally. Choice sends the company record out; your side of the comparison never comes back.
• No tracking in the extension. The extension sends no analytics, usage events, advertising identifiers or error reports. After you allow domain checks, it can ask Choice for the current site’s company record; it also requests aggregate issue-coverage counts, searches when you type in its search box and sends a report when you rate a finding up or down. These operational requests carry no account, device ID or analytics cookie.

When Choice has no record for a domain it returns “unknown” and adds the bare domain to its research queue: the domain, a count of how often it has been asked for and when it was last asked — no address, identifier or page. The most-requested companies are researched next, which is how the directory grows with where people actually go. Cloudflare processes the bare-domain request in transit as Choice’s network and security provider; Choice’s origin does not write the domain to an access log. Full details are in the privacy policy.