Documentation Using the Platform

How to Use a Reach Dog Selling Guide With AI

You do not need prompting experience

Use the category-specific prompt in a selling guide as the starting point. Paste facts from one product record, not a general description of the category. The useful source material includes dimensions, materials, adjustment ranges, ingredients, certifications, compatibility, instructions, limits, testing, warranties, and manufacturer guidance when those fields apply.

Stage 1: Find supported questions

Ask the assistant to compare each buyer question with the supplied product record. Require one of three decisions:

Require the assistant to quote the exact product fact behind every Supported decision. A category association may explain why a question deserves investigation, but it cannot substitute for a product fact.

Stage 2: Draft one piece at a time

Choose one supported question and one channel. Ask for a product-page section, FAQ answer, social post, email, or short-video plan. Supplying the channel, desired length, verified facts, and buyer question produces a more usable draft than asking for generic marketing copy.

Keep the factual product title as the anchor. Put situations, explanations, demonstrations, and comparisons in the rest of the product page and supporting content instead of forcing every phrase into the title.

Stage 3: Check every claim

Paste the draft back into the assistant with the product facts. Ask for a table containing the sentence, evidence type, supporting fact, decision, and safer rewrite.

The allowed evidence types are:

Remove unsupported statements. When a necessary detail is absent, label it MERCHANT VERIFICATION NEEDED rather than asking the model to complete it.

Health, safety, and performance wording

Natural buyer questions may mention a bad back, pain, sensitive skin, safety, durability, or performance. A guide may preserve that language as the shopper’s question. The answer must still describe verified construction, ingredients, adjustments, testing, or documented limits instead of promising a medical, safety, or performance result the product evidence does not support.

A practical first test

Start with one product and one well-supported question. Publish the answer in one visible location, record the baseline available to you, and observe relevant search queries, impressions, qualified visits, engagement, and conversion behavior. Treat the result as evidence for the next change, not proof that one edit caused every movement in traffic or sales.

See these ideas against your own catalog.