Claims and instructions explained

If you sell something where the wording is regulated, you cannot leave it to the AI to decide what your products do. Product rules fix that. They come in two kinds, each on its own tab: Brand → Products → Claims and Brand → Products → Instructions.

Two kinds of product rule

• A claim is what a product may be said to do, be good for or suit — "Vitamin C contributes to the normal function of the immune system." Claims also decide which products your AI recommends when a visitor names a goal, such as "I want to sleep better".
• An instruction is always followed when your AI talks about the products it applies to — a warning, a restriction, a word that must never be used. "Not recommended for children, or for pregnant or breastfeeding women."

A claim always reaches its products through keys. Only an instruction can apply to Every product of this brand instead, which suits wording that holds for your whole catalogue, such as a general disclaimer. There is no "always" or "when relevant" setting any more: an instruction that reaches a product travels with it every time.

Never write a claim about what a product contains ("Contains gluten"): someone asking for gluten-free food would be shown exactly the products with gluten. Put that in an instruction instead.

Keys connect them to products

• Keys — the words a claim or instruction and a product have in common. Usually an ingredient or a product name.
• The text — the sentence your AI reads.

When a product appears in an answer, every switched-on instruction sharing a key with it is handed to your AI as binding wording that overrides everything else it found, overrides anything said earlier in the conversation, and overrides what the visitor asks for. The product's claims travel with it too, as the only statements your AI may make about what it does.

Matching is exact, not clever

A claim or instruction reaches a product when they share at least one key. There is no AI in this matching and no "close enough" — just an exact match on a tidied-up version of each key.

Tidying means: everything is lowercased, accents are stripped, punctuation becomes spaces, and repeated spaces collapse. So C-Vitamin, (Ascorbinsav)! and c vitamin ascorbinsav are the same key.

Synonyms and translations are not matched for you. A claim keyed only on "vitamin C" will never reach a product whose only key is "ascorbic acid". Put every spelling of the same thing — every language, every trade name, every abbreviation — into one claim or instruction.

Where a product's keys come from

A product gets keys from two places, and the product's page shows you which came from where:

• the keys your shop sends with the product, in its rule_keys field;
• the values of any product fields you nominate in the source settings — name your Ingredients field, for instance, and every ingredient becomes a key automatically.

Changing which fields become keys takes effect immediately, with no refresh to wait for.

Claims mode

Claims mode has no switch of its own. It is on as soon as one claim is switched on, and off again when none is. The top of the Claims tab always says which: Claims are on or Claims are off. It is meant for food supplements, cosmetics, medicines and other regulated products.

While claims are on, your AI:

• says what a product does only with that product's own claims — in the same meaning, never stronger, with any condition the claim states, and with nothing added to it ("…, which also matters on a diet");
• says a claim that names an ingredient about that ingredient, as contained in the product ("contains creatine, which …"), never as if the product itself caused the effect;
• gives each claim only to an ingredient that has it, never to a group such as "vitamins", lists only products that have a claim under it, and uses headings that name a kind of product or an ingredient, never an effect such as "immune-boosting";
• still shows products that have no claim, with their facts, but does not say what they do or recommend them for a goal — and never uses your shop's category as the reason;
• shows no products for a goal that no claim covers, says briefly that it cannot recommend a product for that, and asks what else the visitor is looking for — a product the visitor named stays in the answer;
• never says or suggests that a product prevents, treats or cures anything;
• asked what an ingredient or product does when no claim covers it, says once, in one short sentence, that it can't tell what it does in the body — it may still say what it is and what it contains;
• apart from that sentence, never mentions claims or rules. Asked where a statement about what a product does comes from, it says it is the shop's approved product information.

While claims are off, your AI describes your products from their own data. Either way, it never adds general knowledge of its own — see how your AI talks about your products.

The Claims tab at a glance

Under the on/off line, a short checklist keeps count, for example:

• 12 claims switched on
• 40 products have at least one claim
• 8 products have no claim. This line appears only while claims are on: those products still appear in answers, but your AI won't say what they do or recommend them for a goal.

Each claim's row shows Applies to N products — how many active products carry one of its keys — or No product has these keys when it reaches none.

Switching claims on and off

Because one row changes how your AI talks about every product of the brand, the first and the last claim ask before anything changes:

• switching on the brand's first claim — by hand, by approving an import, or by moving an instruction to Claims — asks Switch claims on for this brand?
• switching off, deleting or moving away its last switched-on claim asks Switch claims off for this brand?

Press Cancel and nothing changes. Otherwise, switch each row on or off with the switch in its On column, or tick several and use Switch on selected or Switch off selected. Move to Instructions and Move to Claims carry rows between the two tabs. An instruction set to every product stays an instruction, because a claim always belongs to products through their keys.

How claims find products for a goal

When a visitor asks for products for a goal ("I want to lose weight"), your AI first searches your claims by meaning, then a quick AI check decides which of them really fit what the visitor is after. The chosen claims bring the products that carry their keys, and look-alikes are left out. A product the visitor names is always kept, even when they leave out your brand or the pack size. If it has no claim for that goal, your AI says once that it can't tell what the product does for the goal, gives only its facts, and then shows the products whose claims cover the goal in a group of their own.

Only the claims the check chose travel with the products; when it chose none, each product's claims all travel. Instructions always travel. A new or edited claim is picked up at the next product question — there is nothing to import and no button to press. The check works in written conversations; it does not run on phone calls, when your AI browses a category of your catalogue, or in the answer written after a visitor submits a form. It uses a little AI, billed like the rest of the answer.

Writing a good claim

• The first sentence is the claim; any sentence after it is its condition. Your AI always says the condition together with the claim. For example: "Creatine increases physical performance in successive bursts of short-term, high intensity exercise. The beneficial effect is obtained with a daily intake of 3 g of creatine."
• One claim per rule. Everything after the first sentence is read as a condition, so a second claim in the same row would be said as a condition of the first.
• A nutrition claim is a plain sentence: "The product is sugar-free." — not just "Sugar-free".
• Legal thresholds and sources go into the note. The Note for your team is never shown to your AI, so it is the place for a regulation's entry number, or the minimum amount a product must contain for the claim to be allowed.
• Keys no broader than the claim. A claim that holds only for fish oil gets fish-oil keys, not "Omega-3" — otherwise it also reaches every other product with omega-3 in it.

Adding one by hand

On the Claims tab, press Add claim; on the Instructions tab, Add instruction. You fill in the keys (an instruction can apply to Every product of this brand instead), the text (up to 5,000 characters), an optional Note for your team that the AI never sees, and the on/off switch.

One you type yourself is switched on when you save it, unless you switch it off — a person wrote it, so there is nobody left to review it. Claims and instructions an AI wrote are a different story: see importing from a document.

Where they apply

Every answer a visitor can see: the chat widget, email, Facebook, Instagram, phone calls, and the AI reply suggestions your team sees in the inbox.

They are deliberately not applied where nobody is being answered — the panel that shows an agent what your knowledge holds, and AI social posts.

Two safeguards worth knowing

• Switched-off claims and instructions never reach your AI. Not on any channel.
• A product is never described with only some of its claims and instructions. If there were ever too much of that text for one answer — or simply too much retrieved content overall — whole products leave the answer instead, passages, product card, claims and instructions together.

If a claim or instruction is not doing what you expect, see why a claim or instruction is not applying.