Build a Dedicated AI Assistant Trained on Your Catalog and Brand Voice

1. Write or update a brand voice document: a page describing your tone (casual, formal, playful, technical), a short list of words or phrases you avoid, and two or three example sentences that sound like you
2. Export your product catalog to a spreadsheet or document with SKU, product name, key features, and price
3. Save your return policy, shipping policy, and any other store policy text as a document or PDF

What you should see: Three files ready to upload: a brand voice doc, a catalog export, and a policy doc. Troubleshooting: If you don't have a written brand voice document yet, write five sentences your ideal listing or customer reply would include, and five it never would. That's a working starting point.

1. Log in to chatgpt.com with your Plus account
2. In the left sidebar, click Explore GPTs
3. Click + Create in the top-right corner
4. Switch to the Configure tab
5. Give it a name (for example, "My Shop Assistant") and a short description
6. In the instructions field, paste your custom instructions (see the template below)
7. Under Knowledge, upload your brand voice document, catalog export, and policy document

What you should see: A configured GPT with your name, description, instructions, and three uploaded files listed under Knowledge. Troubleshooting: If the Configure tab isn't visible, confirm you're logged into a Plus account. Free accounts cannot create Custom GPTs.

If you already pay for Pro rather than ChatGPT Plus, a Claude Project does the same job.

1. Log in to claude.ai with your Pro account
2. Create a new Project
3. In the Project's knowledge area, click Add content
4. Upload your brand voice document, catalog export, and policy document
5. Add your custom instructions in the Project's instructions field

What you should see: A Project workspace with your files listed under its knowledge base, ready for any new chat started inside that Project to reference automatically. Troubleshooting: Every chat started from inside the Project inherits the same files and instructions; a chat started outside the Project won't see any of it.

1. Open a new chat inside your Custom GPT or Project
2. Ask it to draft something using your actual catalog: a listing, a customer reply, or a social caption
3. Check whether the tone and details match your brand and your real product info

What you should see: Output that references your actual product names, prices, and policies without you retyping them, in a tone that sounds like your existing listings. Troubleshooting: If the output ignores your catalog details, double-check the file uploaded correctly and that it's in a supported format (PDF, DOCX, CSV, or plain text work reliably in both tools).

1. After a week of real use, note anything it keeps getting wrong: a tone slip, a policy detail it misses, a product fact it invents
2. Update your instructions or knowledge files to correct it
3. Re-test with the same kind of request to confirm the fix worked

What you should see: Output getting closer to what you'd actually publish, with fewer corrections needed each time you use it. Troubleshooting: If it keeps inventing product details that aren't in your catalog file, add an explicit instruction telling it to say "not in my catalog data" rather than guessing when it isn't sure.