Search terms report: paste the CSV into the AI or let it pull the data?

Some people still export the search terms report to CSV, paste it into ChatGPT or Claude, and ask for negatives. Others connect the ad account and let the AI pull the data itself. For a first pass on a new account, pasting is the safer bet because you can see exactly which rows went in. A live pull can quietly cover a different date range or mix in another campaign when the prompt is vague, and you might not catch it until the negatives are already added.

The live connection wins on repeat work, though. Once you trust your rules, pulling the last 30 days fresh each week beats re-exporting by hand. Do you ever let the AI draft the negative list before you've read the raw rows?

I'm not convinced pasting is the safer bet. A CSV without date and campaign columns looks exactly as clean as a live pull, so unless the export includes those fields you can't tell what actually went in. And for e-commerce, a fixed 30-day window can flag terms that were really part of a longer buying path.

Easy to miss on Microsoft Ads, yes. The search query export puts a title block above the column headers, and the date range can sit in that block instead of on each row. If you paste only the table, the period is gone, so check that each row has a date or at least that the header shows the range. For e-commerce I'd also look at conversion date rather than click date, because otherwise a term that converts a week after the click looks dead in a 30-day window.

Good catch on the title block, that would silently drop the date range. A CSV paste also loses the campaign and ad group columns unless you export them, so a term that matches in two campaigns can look like one row. Do you filter to a single campaign before exporting, or pull the whole account and sort it out after?

Pull the whole account and keep the campaign and ad group columns, then make your decisions per campaign. Filtering to one campaign first hides the overlap you most need to see. If a term converts well in your brand campaign and wastes money in a generic one, you won't spot that from a single-campaign export. Add it to a shared negative list and you cut it everywhere, including where it was working.

So the rule I'd use is to read the full account but act at campaign level. Before adding anything, flag every term that shows up in more than one campaign. Then put the negative only on the campaign where the spend is wasted, and keep shared lists for terms that are clearly bad everywhere. It's a bit more manual, but it stops one overlap from quietly costing you a good campaign.

Agreed on acting per campaign. Before you negate anything, though, look at which match type triggered each term in the search terms report. A term that wastes spend under broad match can work fine as exact in another ad group, so the fix might be a match-type change rather than a negative.

Agreed on match type, and I'd go further and make it a rule. If a term converts under one match type and wastes spend under another, fix the match type first instead of reaching for a negative. A negative kills the whole term, including the good traffic it's already bringing in through exact.