The gooder. Standard: full text.
Every principle, red line and requirement, exactly as written.
The gooder. Standard · Version 1.1 · November 2026 · Reviewed every six months.
1. Principles
Every requirement comes back to one of these.
1. Tell people when AI is involved
Customers, clients, job candidates and staff are told when AI plays a real part in what they receive or how they're treated. Anyone who asks whether you use AI gets an honest answer.
2. A person is responsible for everything AI does
"The tool did it" is never an acceptable answer. A person approved the tool, a person checked the work, and a person is responsible for the result.
3. Protect people's information
Personal and client information only goes into AI tools under a business agreement that stops the company behind the tool training its AI on it, and only as much as the task needs.
4. People decide about people
AI can inform a decision that matters to someone. A person who can disagree with it makes the decision.
5. Don't publish anything fake
Nothing invented reaches a client: no made-up facts, fake reviews, fake testimonials or fake people. AI likenesses of real people are only used with their consent.
6. Anyone affected can reach a person and challenge the decision
Staff know how AI is used on them and their work. Anyone affected by a decision AI helped make can reach a person and challenge it.
7. Report and fix AI mistakes quickly
Honest mistakes are reported without blame, fixed quickly and learned from. The people who need to know are told.
2. Red lines
These aren't scored. A business that does any of them can't be certified, however good everything else is. Most go further than the law, on purpose.
A certified business never:
Decisions about hiring, firing, pay, credit or access to a service are always made by a person. Some laws allow AI to make these decisions with safeguards. We don't.
Never on staff or job candidates.
That includes the regulator, the people affected, your clients and us.
When you join Plus or Pro, you confirm in writing that none of these applies to you. If one is crossed, we review it. On Pro, the outcome goes on the public register.
3. Requirements
34 requirements, grouped into five habits. Each requirement carries a tag:
- Everyone: Every business needs it.
- If it applies to you: Only if you use AI in that way.
- From your second year: Needed once you've been with us for 12 months.
- Core: Nine requirements everything else depends on. A gap in one always has to be fixed before you can be certified.
On Plus, we write the policies and procedures that put each one in place, and keep them current.
Habit 1: Know which AI tools you use
You have a list of every AI tool in use, each one was approved by a named person, and you know what information it can reach.
One named person approves tools and hears about problems, and is senior enough to say no. Teams of ten or more also name a deputy.
Every AI tool and feature in use, including AI built into software you already pay for, approved before anyone uses it for work. What data each touches, which plan you're on, who approved it and when. Updated within 30 days of any change.
The plan, the data agreement, whether it trains on your data, how long it keeps it, where it's held, who else handles it, and how you'd leave. Accounts are secure: two-step login on, no shared logins. Checked again every year.
People get AI access once they're trained, and lose it the day they leave.
A short, recorded review each quarter asks what's new and what's changed. The policy gets a full review every year.
Before AI that can search your files or email is switched on, you review what it can reach. You repeat that every year.
AI that can send, pay, delete or change records needs someone to confirm each of those actions.
Habit 2: Train your team and set clear rules
Everyone knows what they can and can't put into AI, and your AI policy is how you actually work, not a document in a drawer.
Signed by everyone who uses AI for you, contractors included, with an approved tool list that matches your register.
Information sorted into public, internal and restricted, with the tools allowed for each.
Everyone is trained before they use AI for you, with a dated record. Everyone completes chapters 1 to 7 of the course, managers also complete chapters 8 to 10, and the AI owner completes all ten. New starters complete the new-starter module before they get access to AI tools.
Staff and freelance contracts cover confidentiality when using AI. Freelancers assign you the rights in the AI-assisted work they do for you.
New payees and changes to bank details are checked on a number you already hold, and approved by two people.
Refreshed every year, and within 60 days of a big change to your tools or rules.
Habit 3: Have a person check AI's work
AI's work is checked by a person before anyone relies on it, and decisions about people are made by people.
A written rule: a sense check for internal drafts, a full read for anything that goes out, and sources checked by a second person for advice, figures and legal points. In businesses of two or more people. Sole traders check every fact against its original source, and keep a note.
Where AI helps produce client work or content you publish, facts, figures and citations are checked against the original, with a note kept. It doesn't copy someone else's work, imitate a living artist's style, or clash with someone's trade mark.
Any decision that significantly affects someone is made by a person with the knowledge, time and authority to disagree with the tool. Their reviews are recorded.
A data protection risk check before using AI with people's information at scale, and a full impact assessment where that check says so.
Anyone affected by an AI-informed decision is told, and can ask a person to review it and challenge it.
Evidence of bias testing from your vendor. From your second year, a yearly check of outcomes across groups, where you have enough decisions to compare.
Customer-facing AI is tested against your real policies before launch and after every change, and the tests are recorded.
Habit 4: Tell people when you use AI
Customers, clients and anyone whose information you use know when AI is involved, and can reach a person.
Customers, staff and candidates are told which AI processes their information, which companies are involved, where it goes, how long it's kept, and how to complain.
An AI clause in your client terms, any client restrictions recorded where the work happens, and existing clients told.
Easy to find, such as a link in your website footer, and in line with your register.
Staff are told how AI is used on them and their work, including any monitoring, and are consulted before AI changes how their work is judged.
Chatbots and assistants say they're AI from the start, and a person is easy to reach.
AI-generated images, video or voice showing people are labelled as AI-made.
If you have customers or users in the EU, they're told when they're dealing with AI, and your team's AI literacy is supported.
Habit 5: Deal with AI problems properly
When something goes wrong, you report it, fix it and learn from it.
Anyone can report an AI mistake to the AI owner within 24 hours of noticing, and honest mistakes are never punished.
Each one is recorded and checked against any legal deadline for reporting a data breach, such as 72 hours under UK and EU GDPR, or under your state's breach law.
A clear way to complain about how you use people's information and AI, acknowledged within 30 days.
Each incident notes what caused it and what changed as a result.
You tell gooder. within 10 working days of a significant incident or a big change, such as starting to use AI in hiring.
Your AI use is described accurately to your insurer, and any AI exclusions in your policy have been read.
A 30-minute walkthrough of a realistic incident, so the first time isn't the real one.
4. How it's checked
You don't send us a pile of evidence. Each requirement is checked in one of three ways.
The weekly scan
Every week we read your website and the tools on it, the way your customers do. The scan checks what anyone can see: copy and images that read as AI, reviews and people that look invented, claims about AI you'd need to back up, whether your chatbot says it's AI, whether your privacy notice covers the AI tools on your site, and the AI terms of those tools.
Covers 1.2 and 1.3 for the tools on your website, 3.2, 4.1, 4.5, 4.6, and red lines R5 and R6.
Your gooder. account
What you do with us is recorded as you do it: your AI policy written, approved and signed by each person, everyone's training and check results, who joins and leaves your team, and the declarations you make, such as your chatbot telling visitors it's AI.
Covers 1.1, 1.4, 2.1, 2.2, 2.3, 2.6 and 4.5, and your red lines declaration.
One human review a year (Pro)
Once a year, one of our team goes through the rest with you and asks the three questions for each requirement: is there a record, does it happen in practice, and does your team know about it?
Covers your tool list and quarterly reviews, tool checks, contracts, payments, decisions about people, chatbot testing, client terms, your "How we use AI" statement, and your incident log, complaints and yearly practice.
5. What happens when something's wrong
Your status and badge follow your latest weekly scan. They can go down, and they come back when you fix things.
- A high-risk issue gives you 14 days to fix it. After that, your status changes to Under review, and so does your badge.
- If it's still unfixed 30 days after that, the badge comes down and your Certified listing is removed.
- A fix counts once it holds for two weekly scans in a row, or once we've confirmed it. Then your status and badge come back. There's nothing to reapply for.
- A lapsed policy or lapsed training counts as a high-risk issue, the same as anything the scan finds.
- If a declaration you made turns out not to be true, such as your chatbot not saying it's AI, we may end your certification.
- If a red line is crossed, we review it. On Pro, the outcome goes on your public register entry for 12 months, whether it was put right or your certification ended. Two people at gooder. agree every outcome before it's published, and you can ask for it to be looked at again.
The statuses on the public register: Monitored, Policies in place, Working towards certification, Certified, Under review, Certification ended.