An AI product persona for ecommerce is a short brief that tells a store assistant how to speak about one item, which shoppers it is meant to help and which catalog facts it can rely on. To write one for WooCommerce, start with the product’s purpose, the buyer it suits and the details the shop can verify. Then choose a voice and specify what the assistant should do when a shopper asks for a fact the store hasn’t recorded.
For a leather messenger bag, the persona can sound like a concise, practical shop assistant helping someone choose a work bag. Its tone can make a sleeve measurement easier to understand, but it can’t turn that measurement into a promise that a laptop will fit. The catalog supplies the facts; the persona decides how to explain them. A good brief makes that distinction clear enough that an editor can test it in a conversation.
Start with the buying decision, not the character
Picture a commuter comparing two messenger bags. They aren’t asking whether the bags have an appealing personality. They want to know whether a laptop in its case will go through the opening, whether the black option they’ve switched to is in stock and what happens if they need to return it. Those questions tell you more about what the persona needs than a list of adjectives like friendly, premium and confident.
Write one sentence that describes the decision the assistant should help with. For this example: Help a commuter decide whether this messenger bag fits the items they carry and whether their chosen option is the one they want to order. This gives the assistant a reason to ask which dimension the shopper measured on the laptop case. It also gives the editor a reason to cut a colourful line about the bag that does nothing to help with that decision.
Next, note where the assistant should look. The product description supplies item-level details such as intended use and the stated material. A documented measurement needs a label that says what was measured. The variation record is where the editor checks the relevant colour, price and stock. The shop’s current policy is the place to check shipping and returns. If the description calls a dimension “width” without identifying what was measured, fix or flag that ambiguity rather than asking the persona to write around it.
Keep a short list of questions the available facts cannot settle. For this bag, the important one is whether the opening admits a cased laptop of a given size when only the inner sleeve width is known. Put that question in the brief alongside the known details. Otherwise, an editor may see a fluent answer containing the correct number and miss that the assistant used it to answer a different question.
Write a compact persona brief for the messenger bag
The following is an example brief, not a live listing. The 34 cm measurement belongs to the review scenario below. Before using the brief on a real product, replace example details with the shop’s verified records. Use the exact material wording and the colour labels the store actually publishes; don’t fill either field from the example’s product name.
- Item and purpose: A leather messenger bag designed for commuting. Help shoppers decide whether it suits the items they plan to carry.
- Shopper: Someone comparing work bags, often with a laptop already in a case. Ask for the outside dimensions of the case if the answer depends on them.
- Voice: Concise and matter-of-fact. Use everyday terms such as “inside the sleeve” and “across the opening.” First person is optional in one or two short replies if it fits the brand; switch to direct language when explaining a limitation.
- Product facts: Use the description and recorded measurements for construction and dimensions. In this example, the listed inner sleeve width is 34 cm; the opening width isn’t listed. Use the exact catalog specification for the material.
- Option facts: Name only the specified colours. For a reply about colour, price or stock, check the exact variation the shopper means. Ask which one they mean if the conversation doesn’t identify it.
- Policy and gaps: Refer to the store’s current shipping and returns policy. If a fact needed for the decision is missing, name it plainly and offer a check only if the shop can actually arrange one.
This brief is short enough to review next to the product record. Keep changeable details such as stock out of the voice instruction. For example, “mention that black is available” is a poor standing instruction: it turns a variation fact into something the persona might repeat even when it no longer applies. “Check the black variation before answering” tells the assistant what to verify without prescribing the result.
Test the voice after you’ve checked the facts. “My listing gives the sleeve width as 34 cm” and “The listed inner sleeve width is 34 cm” can both work. The first gives the product a light first-person voice; the second is more neutral. Pick the one that feels natural for the shop. Neither gets to add “so you’ll have plenty of room.” That would be a new claim about fit, not a change in style.
Apply four product truth rules
Use these rules when editing the brief and reviewing replies. Each rule identifies a different place where a plausible answer can drift away from what the shop knows.
1. Quote the inner measurement and explain where it was taken
Say “34 cm across the inside of the laptop sleeve,” not “34 cm of laptop space.” The first describes the recorded dimension in the example. The second could be read as a usable limit for any device or case. If the shop measured across the opening, call it the opening width. If the record doesn’t say how the measurement was taken, don’t invent details about the tape position or whether the sleeve was stretched.
Check the shopper’s dimension too. “A 34 cm laptop with a case” could mean the laptop alone is 34 cm wide or the whole cased laptop is 34 cm wide. Ask a short clarifying question if that distinction matters. Even when the shopper confirms an outside width, compare like with like: an inner sleeve width doesn’t establish the width of the passage into it. A matching pair of numbers is not, on its own, a fit check.
2. Describe the material using the exact product specification
If the catalog says “leather,” use that wording rather than upgrading it to “full-grain leather” or “premium leather.” If the specification names the material for only one part, keep the claim tied to that part. A statement about a leather exterior, for example, does not specify the lining or strap. If a shopper asks about a finish or treatment that the listing doesn’t identify, say that it isn’t specified.
Ordinary language is still useful. A reply can say “The listing specifies leather for the exterior” instead of copying a field label that makes little sense outside the catalog. To review a rewrite, underline the words that describe the material. If the rewrite includes a grade, finish or component the source doesn’t name, remove the extra claim. An inviting voice doesn’t need a more flattering specification.
3. Check the selected variation for price, colour and stock
A shopper can start by asking about one bag option and then switch to another. Treat “What about the other black one?” as a possible change of selection, not as permission to repeat the answer for the first option. Identify the variation by its catalog label or another distinguishing option the shopper can confirm. Then check the price, colour and stock for that record before answering.
If two variations share a colour label, “black” alone may not identify an orderable option. A useful question is “Which black option do you mean?” followed by the actual distinguishing labels if they’re available. Don’t assume a colour change also changes the sleeve measurement, but don’t assume the measurement carries over either. Check whether the dimension belongs to the whole product or to the particular variation.
4. Retrieve shipping and returns from the store’s current policy
Keep shipping and returns out of the persona’s standing sales lines. A shopper asking “Can I return it if my laptop doesn’t fit?” is asking about policy, not about the bag. Use the current policy and preserve any conditions that matter to the answer. If the relevant details aren’t available in the conversation, don’t substitute a remembered policy or promise that a return will be accepted.
Answer the product and policy parts separately when a shopper asks both at once. Explain why the fit is uncertain, then give the relevant return terms if they can be checked. A return policy cannot establish whether the bag fits, and an unanswered measurement question does not establish whether a return is permitted. Keeping the two answers distinct helps the shopper decide whether to wait for a measurement before ordering.
Show what happens when a specification is missing
Include a missing-fact dialogue in the test sheet, not just a list of correct product statements. A well-grounded reply should explain what the known measurement means, name the missing measurement and avoid mistaking an offer to check for a completed check. Here is the example dialogue for a shop that can ask its team to measure the bag:
Shopper: “Will this fit a 34 cm laptop with a case?”
Product chat: “My listing gives the sleeve width as 34 cm. The opening width is missing, so I can ask the team to check it with the case in mind.”
The useful part is not simply that the reply declines to say yes. It explains why the number it has doesn’t settle the question. It also says “I can ask,” not “I’ve asked” or “the team will reply today.” If the shop has no way to pass along a measurement request, change the last clause to an available next step or say that fit can’t be confirmed from the listing.
If the conversation continues, a useful next question is “Is 34 cm the outside width with the case on, or the width of the laptop alone?” The assistant can ask for the case’s other outside dimensions if the shop needs them for a fit check. It shouldn’t make the shopper provide details without explaining why they matter. It already knows which part of the product record is missing, so it can name that gap first.
Run the same prompt against a second example product: a larger bag whose measured opening is listed. This time, expect the reply to name that opening measurement and compare it with the outside width of the case if the shopper has confirmed it. Don’t require a blanket “yes” just because the opening is wider. The answer should say what that comparison establishes and flag any other dimension the recorded facts don’t resolve. If both products get the same “opening width is missing” answer, the assistant isn’t using the difference between their records.
Add a mid-conversation switch: “Actually, what about the other black variation?” Expect the assistant to resolve which version the shopper means before making a new price or stock claim. If the shopper still wants a fit answer, the assistant should check the scope of the measurement rather than carrying it over silently. This test catches a reply that was accurate for the first option but becomes wrong after the subject changes.
Keep voice, the accepted cart action and payment separate
Voice selection is an editorial decision. Choose how the assistant explains a verified detail, asks a follow-up or recommends comparing two options. A neutral voice might say “Check the case width before choosing this size.” A light first-person voice might say “My sleeve is listed at 34 cm wide inside.” The wording changes, but the product facts and the shopper’s selection do not.
The accepted cart action is a separate step. For a test, define exactly what the shopper is accepting: adding the identified bag variation to the cart. “Is the black one available?” is a question, not an instruction to add it. “Yes, add that one” only identifies a usable action if “that one” already refers to a specific option. Even after the shopper accepts, a reply saying the item was added should reflect the actual outcome, not merely the shopper’s intent.
Payment is distinct from both the chat recommendation and the cart action. Don’t write “You’ve bought it” or “Payment received” based on an accepted add-to-cart action or a successful cart update. The persona brief can tell the assistant to acknowledge the cart step in the chosen voice. It cannot treat that acknowledgement as evidence that checkout or payment has happened.
Label any mock chat preview as illustrative. A preview can show the proposed wording and the intended sequence of a conversation. It does not verify that a live store resolved a variation, added an item or processed a payment. Test the language in the preview; check actual actions separately in the store workflow before describing them as completed.
Separate workflow descriptions from listing evidence
For Senoji Baldinė, distinguish the owner-described workflow from facts you can verify in the public product listing. Attribute an account of how the shop works to the owner. Use the listing to substantiate only what it actually states about the product. A workflow description can explain how someone handles a question, but it doesn’t, by itself, establish a material, a dimension or the availability of a particular option.
Keep that distinction visible in the editorial review. For a workflow statement, record who described it and use wording that makes the attribution clear. For a product claim, note the field that supports it and whether that field applies to the whole item or a specific option. If the listing is silent on a detail, leave it unverified for the purposes of the article. Don’t use the messenger bag dialogue as evidence of how Senoji Baldinė’s store behaves or of what its public listing contains.
This is also a useful rule for a shop’s own brief. Suppose a team member says they can measure a sleeve opening on request. That can support an offer to ask for a measurement if the shop has a process for it. It cannot supply the missing width. Record the result in the product data once the team has measured it and confirmed the option it applies to. Until then, the answer should remain explicit about what the shopper cannot verify.
Review the replies and measure what helps shoppers decide
Prepare a small test sheet before publishing. For each prompt, record the active product or variation, the expected source, the fact the assistant may state and the claim it must not make. Include a known sleeve dimension, an unspecified material treatment, the cased-laptop question, a switch between black options, a return-policy question and an accepted cart action. Put the larger-bag dialogue beside the first fit test so an editor can compare the two replies directly.
Review individual claims as well as whole answers. A response can quote the sleeve width correctly and still end with an unsupported fit promise. Conversely, an answer that contains no false claim can be unhelpful if it says only “I don’t know” when the assistant has a relevant dimension to share. Mark both kinds of failure. For each one, identify whether the fix belongs in the persona brief, the product record or the shop’s handling of the next step.
- Claim support: Count objective claims that match the relevant record, and note those with no source or with the wrong scope. Look for small additions such as “plenty of room” that change what a correct measurement implies.
- Missing-fact handling: In tests with an incomplete record, check whether the reply names the missing detail, shares the useful known fact and avoids promising an unavailable check.
- Variation accuracy: Count how often replies after a switch use the intended variation. Record a correct clarifying question separately from a confident answer about the wrong option.
- Decision progress: Note whether the shopper gets a usable comparison, knows which measurement to provide or knows what the shop needs to check. Track repeated questions about the same missing specification as candidates for catalog improvements.
- Action accuracy: Log accepted cart actions, confirmed cart outcomes and payment confirmations separately when testing the workflow. An accepted action isn’t a confirmed update, and a confirmed update isn’t payment.
Use the results to make the smallest effective edit. If a reply says “full-grain” when the listing only says “leather,” tighten the material rule and rerun that prompt. If shoppers repeatedly ask about a missing opening width, measuring and publishing it is more useful than polishing the explanation of why it’s missing. If the wrong black option keeps appearing, fix how the selection is identified before changing the tone. Retest the affected prompts when catalog details or store policies change. The persona is ready when it can help a shopper take the next step without making the facts sound more certain than they are.
Related guides
- Seven conversational commerce examples for online stores
- Guided selling for shoppers who arrive with a problem
