Whichever you load into your factor table. Common starting points: DEFRA UK government factors (updated yearly), EPA eGRID for US electricity, IPCC AR6 for global warming potentials, Ecoinvent for LCA. The bot quotes the source for every factor in its reply, so users see which dataset drives the number and you can swap factors as standards evolve.

For directional product footprints, yes. Store per-category LCA assumptions (kg CO2e per kg of cotton, per kWh of compute, per kg of packaging) and the bot can sum them per product. For audit-grade Scope 3 reporting, the chatbot is a triage tool that points users to a formal LCA process, not a replacement for a full assessment.

Yes, but with care. The bot can estimate Scope 1 (fuel combustion), Scope 2 (purchased electricity), and high-level Scope 3 categories from input data. Reporting-quality numbers still need a full inventory process. The bot is best used as a starting estimate, a learning tool for non-experts, or a quick check between formal reporting cycles.

EPA eGRID factors per grid subregion are stored keyed by postcode. The bot looks up the user's subregion and uses the right kg CO2e per kWh. Renewable-energy credit (REC) purchases can be reflected with a market-based factor if the user provides supply mix details. Both location- and market-based methods are supported.

Not directly. The bot links to your offset purchase flow (Stripe Checkout, Snipcart, or a WooCommerce product) with the right tonnage pre-filled. Keeping payment out of the chat keeps the bot focused on estimation and the purchase flow on your existing tax and compliance plumbing.

Most authoritative sources (DEFRA, EPA, IPCC, Ecoinvent) publish annually. Schedule a yearly refresh task: import the latest CSV into your factors table, update vintage years on the sources, run a few smoke-test conversations to catch regressions. The bot uses the latest data on the next request, no redeploy.

For consumer-facing tools a single point estimate works. For corporate or technical audiences, the bot can return a low-mid-high band per line item, sourced from the factor's uncertainty range. The reply gets longer, so reserve banded answers for users who explicitly ask for them or for B2B-facing scoping.

The chatbot is not an audit substitute. It is a fast, transparent estimation tool grounded in your stored factors. Many sustainability teams use it for early-stage scoping, employee engagement, and consumer-facing communication, then run formal Scope 1-3 inventory through dedicated software for CSRD, SEC, or other regulated disclosure. The bot's logs can feed into that workflow as a starting point.