AI assistant: how it works, safety and privacy

Affiliate Buddy is an AI assistant, not a person. It runs on an AI model, Claude Haiku 5.5 by Anthropic. The chat always shows an AI label, and before your first chat it shows this notice:

I'm an AI and can make mistakes. Don't share personal info.

Every new chat then starts with a short reminder: "AI can be wrong. Don't share personal info."

It only uses FindAffiliates listings and pages, such as program profiles, categories, the Program Finder and the pricing page. It does not browse the web, and it does not use other directories, networks or review sites.

It can be wrong. Program terms change, and a listing can be out of date. Check the program's page on FindAffiliates and the program's own terms before you rely on an answer.

It stays closed until you open it, and it does not remember past chats. Each chat starts fresh.

It can

• Find programs listed on FindAffiliates by name or category.
• Explain the terms a listing shows, such as the commission and the cookie length.
• Show your Program Finder match and point you to free tools such as the commission calculator.
• Explain what a free account and PRO include, and show companies how to list a program.

It won't

• Promise or predict earnings, approvals, traffic or results. When asked, it says: "Results vary; no earnings are guaranteed."
• Give personal financial, legal, tax or business advice.
• Apply to programs, contact companies, or change anything in your account.
• Link to other websites, or help with topics that are not about affiliate programs and FindAffiliates.
• Pretend to be a person or a character, role-play, or ask about your feelings or personal life.

How much you can use the assistant is measured by how much the AI reads and writes, not by the number of messages: short questions use less, and long chats and searches use more. A free account gets a daily allowance, you get a smaller one without an account, and PRO gets 3× a free account's daily allowance, with a monthly cap. The chat menu shows how much of today's allowance is left. Without an account or with a free account, questions also count toward a shared daily limit for each network, such as a school or office connection. The assistant can also pause for a while when it is busy or when its daily or monthly budget is used up.

Before the first chat, the assistant asks one question: "What year were you born?" The field starts empty, and there is a "Prefer not to say" button. We keep only an age range, never the birth year. Because we only ask for the year, we use the youngest age the answer could mean.

This is a self-declared check, not age verification. For 30 days after an answer, a new answer can only make the settings more protective. If you are signed in, that also holds on your other devices and after you delete the assistant's cookie, because the age range is kept with a one-way hash of your account ID for those 30 days. Without an account, clearing this browser's cookies starts the age question over. Visitors without an account answer the age question first and then complete a quick human check.

• Under 13: the assistant is not available. It shows a short message with links to the guides and the home page, does not accept or send any messages, and does not ask again in that browser (or, for a signed-in account, on other devices for 30 days).
• 13 to 17, or Prefer not to say: no upgrade or PRO prompts, messages of up to 500 characters, social media handles removed along with other personal details, stricter checks for grooming, and a reminder to take a break after 60 minutes of chatting. For questions about money, payments, taxes, contracts or joining a program, it says that many programs require members to be 18 or older and suggests involving a parent or guardian.
• 18 or older: messages of up to 800 characters. A reply may mention once what a free account or PRO adds, when that is relevant to the question.

Plan limits are the same in every mode. Many affiliate programs require members to be 18 or older, and FindAffiliates is built for adults who run businesses, websites, newsletters or channels.

Every message is checked before it reaches the AI model, and every reply is checked again before you see it.

If someone may be in crisis

The assistant checks every message for common words and phrases about suicide, self-harm, abuse, bullying and people asking for photos, secrets or meetups. When it finds them, it shows a fixed message with free helplines instead of an AI reply, even when your daily limit is used up. This check can miss some wording, so the helplines below are always available.

The crisis message says that the person deserves support from a real person and does not have to face it alone, that the assistant is an AI, where to find free help right now, and to call the local emergency number if they might act on these thoughts or are in danger. It also suggests talking to someone they trust. The AI model is told to show the same message when it notices that someone is struggling, and the assistant's limit and pause messages carry a link to free helplines too.

Bullying, grooming and sexual content

Messages about being bullied, or about someone online asking for secrets, photos, contact details, a meetup or a move to another app, get a fixed message that suggests telling a trusted adult and points to findahelpline.com. If someone says they were asked for nude photos, the message says it is not their fault, that they do not have to send anything or keep it a secret, and how to stop replying, block and report the account and get help. Requests for sexual content get a fixed refusal, and anything that sexualizes minors is always refused. Messages the check catches are not sent to the AI model.

Checks on every reply

Replies are checked for unsafe content, for promises of earnings and for links. The assistant can only link to FindAffiliates pages, so any other link is removed, and an unsafe reply is replaced with a fixed message.

We count how many times each safety message is shown each day, by country, with no information about who saw it. We do not see chats unless someone reports a reply, and we cannot contact anyone through the assistant.

This AI assistant may not be suitable for some minors.

Please don't share personal information in chat, such as your name, address, phone number, email or passwords.

What is sent

When you send a message, our servers receive it together with the recent messages of the same chat, the page you are on, your plan, your interface language and your age range. The AI model is told only what kind of page it is, such as a program page, a category or the pricing page. If you are signed in and ask about the Program Finder, the assistant can also use your saved Program Finder answers.

Before anything leaves our servers, we try to remove things that look like personal information: email addresses, phone numbers, names and schools people mention, usernames given with their app (such as "my snap is"), payment card, ID and bank account numbers, IP and street addresses, and the query part of links. For users aged 13 to 17 and users who prefer not to give their age, we also try to remove other social media usernames, such as @handles. This is automatic and can miss things, so please don't share them.

Who receives it

To write a reply, we send the cleaned message, the recent chat, those details and the FindAffiliates listing data needed for the answer to Anthropic, the company that makes the AI model. We also send a code made with a one-way hash of your account or browser ID, which helps Anthropic detect abuse. It contains no name or contact details. For users aged 13 to 17 and users who prefer not to give their age, this code changes every day. Anthropic's policy is to delete these messages within 30 days, unless it must keep them longer, for example to enforce its usage policy (for up to 2 years) or because the law requires it. Under its commercial terms, Anthropic may not use these messages to train its AI models.

In Simple mode, messages are processed on our servers to pick an answer and are never sent to an AI company. Only counters are kept.

What we keep, and for how long

Your chats

Not stored on our servers. A chat stays in your browser tab until you close or reload it, or start a new chat.

Usage counters

Usage and message counts linked to a one-way hash of your account ID, a random browser ID or your network address. Most are deleted within 3 days; monthly counters within 33 days.

Reported replies

30 days, then deleted. A report holds the reply, the reason, an optional note and, only if you choose, the message that led to it, with personal information removed, plus your plan, age range and language. It has no account ID, browser ID, IP address or cookie.

The fa_ai cookie

Up to 180 days, or until you delete your AI settings. It holds a random browser ID, when your human check expires (12 hours after you pass it) with a code for the network you passed it on, and your age range with the date you answered. Never your birth year. It is sent only with the assistant's own requests.

Age setting

For people aged 13 to 17, under 13, or who prefer not to give their age: the age range and the day its 30-day lock ends, kept with a one-way hash of your account ID or random browser ID, then deleted. Under-13 answers are kept this way only for signed-in accounts. Never your birth year.

Storage on this device

fa_ai_hidden (you hid the assistant) and fa_ai_seen (its one-time greeting played) stay until you clear site data. fa_ai_youth (an age setting for under-18s, or for people who prefer not to say, is running, so the site's discount offer stays hidden) holds only the date that setting ends. fa_ai_nudged (the greeting already ran in this tab) and fa_ai_session_start (the break reminder clock) end when you close the tab. fa_ai_mode (you chose Simple mode, without AI) stays until you switch back, delete your AI settings or clear site data. None of them holds chat text.

Daily totals

Counts of replies, reports and safety messages shown each day (safety messages also by country), with no identifiers. Deleted after about 13 months.

Visitors without an account complete a Cloudflare Turnstile check after the age question and before their first message. Cloudflare processes technical signals from the browser to tell people from automated traffic. A passed check lasts 12 hours on the network where it was completed.

If a message suggests that someone may be in crisis, we use the country your connection comes from, as reported by our hosting provider, to show local helplines. The country is kept only in the daily count of safety messages, never with anything that identifies you.

We do not use assistant conversations for advertising or profiling, and there are no ads in the chat.

The Privacy Policy has the full details.

Hide it on this device

You can also hide it from the chat: open the chat menu ("Chat options") and choose "Hide Affiliate Buddy". Hiding works in this browser on this device. To bring it back, use the switch above. The "AI assistant" link in the site footer leads to this page.

Delete my AI settings

In the chat menu, choose "About and safety", then "Delete my AI settings". This deletes the assistant's cookie on this device. Your age range stays until 30 days after you answered, so the age check keeps working, and so does its server-side copy described above.

Report a reply

Use "Report this reply" under any reply, or "Report a reply" in the chat menu. Pick a reason and, if you like, add a short note. Including the message that led to the reply is optional and stays off unless you tick it, and it is never included for a safety message. Reports are kept for 30 days and reviewed by the FindAffiliates team to make the assistant safer.

Start over

"New chat" in the chat menu clears the current chat from the tab.

FindAffiliates is built for adults who research affiliate programs, but teens may find it too. Here is what you can do.

• Hide it on your child's devices. There are no parental-control accounts. Use the switch on this page, or "Hide Affiliate Buddy" in the chat menu, in each browser and on each device your child uses. Clearing site data or using a private window shows it again, and anyone using that browser can turn it back on.
• Ask us. For questions, worries, or to have a report deleted, email contact@rekonmedia.com. We cannot see chats, so include the reply text or roughly when it was sent if you can.

Talking with teens about AI

• An AI can sound sure of itself and still be wrong. Check facts on the program's own page.
• Keep personal things private: real name, address, school, phone number, passwords and photos.
• No one can promise affiliate earnings. Be careful with anyone who does.
• Many affiliate programs require members to be 18 or older. Contracts, payments and taxes are for an adult to handle.
• If someone online asks for secrets, photos, a meetup or a move to another app, tell a trusted adult.

This AI assistant may not be suitable for some minors.

FindAffiliates is a general-audience website built for adults, and it is not directed to children under 13. In line with the US Children's Online Privacy Protection Act (COPPA), the assistant collects no personal information from a user whose age answer shows they may be under 13, apart from the identifiers described below, which support its internal operation only.

The age question before first use is how we find out, and it comes before the human check, so an answer that shows a user may be under 13 is never sent to Cloudflare. If the answer shows that the user may be under 13, the assistant is not available: it does not accept their messages or send them anywhere, creates no browser ID and no usage counters of its own for them, and keeps only the age range in a cookie on that device so the check keeps working. If the user is signed in, the age range is also kept with a one-way hash of the account ID for 30 days, so the check works on their other devices.

To prevent abuse, setup requests such as the age question are counted for everyone per network, with a one-way hash of the network address, and these counts are deleted within 3 days. Identifiers like this are used only to support the internal operation of the assistant, such as security and rate limits, and never for advertising, profiling or contacting anyone.

If you believe a child under 13 has shared personal information with us, email contact@rekonmedia.com and we will delete it.

Questions about the AI assistant, or a request to delete a report, can be sent to contact@rekonmedia.com.

See also the Privacy Policy and the Terms of Service.