FAQ

Enough that someone unfamiliar with your store could write one product description from it and you would recognise the result. In practice that is a few paragraphs covering personality, tone, do’s and don’ts, restricted claims, and vocabulary. See Writing brand guidelines.

Yes. Paste an excerpt from a product page you are happy with and say “sound like this,” then add your do’s, don’ts, and vocabulary rules. That is enough to start.

Brand voice applies to everything you sell. Additional context applies to this batch. If a statement would still be true for a completely different set of products, it belongs in brand voice.

No. Each purpose pulls the writing toward a different reader, and the same seven fields have to serve all of them. Two related purposes produce sharper output than four unrelated ones.

Terms shoppers actually search for, taken from site search logs, Search Console, or ad query reports. Keep the list short — long lists produce copy that reads like a keyword list.

The limit is 250. Start with 25 to 50 in one category to validate your inputs, then scale. Tuning on a small batch costs far less than discovering a problem across 250.

You can, but your context then has to stay general, because it is applied to every product in the batch. Specific context on a mixed batch produces output that fits some products and not others. See Planning your batches.

No. A product counts as one entitlement whether it is simple or has variants and modifier options.

Yes. Products are scored on accuracy, consistency, and brand adherence, and each score comes with the specific issues behind it. You set a minimum per factor before the run, and products below it are held for your review. See Understanding quality scores.

Usually your brand guidelines are too general. A score reflects the problems the reviewer checks for — unsupported claims, contradictions, guideline breaches — not whether the copy suits your brand. If output is clean but flat, add specificity to your guidelines rather than raising the threshold.

It writes from your product data and does not fabricate facts beyond it. The risk is highest on sparse records, so check your thinnest products closely and confirm that any specific claim is supported by the record.

Stop editing and change the input. Repeated corrections mean the brand voice, context, or source data is wrong. See Deciding what to change.

No. Applying overwrites the existing field values permanently, which is why backing up your catalog beforehand and applying a small selection first are both worth doing.

No. That is attribution, which is not part of data enrichment today and is on the roadmap. Enrichment generates copy and will not infer facts your data does not contain.

Baseline first, enrich one category, leave a comparable one alone, then compare the two after about four weeks on organic sessions, conversion rate, and revenue per visit. See Proving enrichment works.

On-site behaviour changes as soon as content is live. Organic search takes weeks, because indexing and ranking do not respond immediately. Answer engine visibility is the slowest and least precise to measure.