ChatGPT for Google Ads keyword research: a practical workflow

What ChatGPT can and cannot do for keywords

ChatGPT is good at generating seed lists, clustering themes, and writing match-type variants of a phrase you give it. It does not have live search volume data. It cannot tell you what a click costs in your vertical today, and it has no visibility into your existing account structure.

That gap is exactly where SpendSignoff fills in. Connect ChatGPT to SpendSignoff over MCP, and the operator can read your actual impression share, wasted-spend queries from your search term report, and the keyword gaps between your current ad groups — then use that data to make ChatGPT output immediately actionable.

The practical workflow: use ChatGPT to expand themes and generate match-type variants, then feed those variants into SpendSignoff as a proposed keyword batch. The operator produces a propose_change draft with the structured list. You review the diff in the SpendSignoff approval queue and approve.

Step-by-step keyword workflow

This workflow runs entirely inside Claude or ChatGPT connected to SpendSignoff. You start with a brief and end with approved keywords in the live account.

Match type strategy by funnel stage

Broad match with Smart Bidding reads signals you cannot see. Use it for top-of-funnel terms where Google has enough conversion history to optimize — typically campaigns with more than 30 conversions per month. Below that threshold, broad match dilutes your signal.

Phrase match is the workhorse for mid-funnel. It respects word order, so "running shoes for flat feet" does not serve on "flat feet shoe running". That containment is worth the reduced reach when your offer is specific.

Exact match is the only type where you know precisely what you are buying. Use it for high-value brand-adjacent queries and competitor terms where intent is unambiguous. Expect higher CPCs — the precision costs something.

A common mistake is applying the same negative lists across all campaigns. SpendSignoff lets you scope negatives at the campaign level, so your brand campaign can remain open to broad queries while your product campaign excludes them.

What SpendSignoff reads to improve keyword decisions

Before drafting any keyword change, the operator reads these data points from your live Google Ads account:

• Search term report — actual queries that triggered your ads, with impressions, clicks, cost, and conversions per query.
• Quality scores per keyword — expected CTR, ad relevance, and landing page experience, which predict your effective CPC ceiling.
• Impression share by campaign — where you are losing to budget versus losing to rank, which tells you whether to add keywords or increase bids first.
• Ad group keyword density — how many active keywords each ad group is carrying; groups over 20 keywords usually benefit from splitting.
• Conversion lag by keyword — some keywords close in a day, others in 30 days; bid decisions need the right attribution window.

Negative keyword maintenance as an ongoing loop

One-time negative lists decay fast. Search behavior shifts, new modifiers emerge, and products change. The SpendSignoff autonomy loop runs weekly by default and flags new high-cost zero-conversion queries as candidate negatives. It does not add them automatically — it drafts them into the approval queue for your review.

This "propose, do not commit" posture means you see the pattern accumulating over time. If the same irrelevant modifier keeps appearing in different campaign contexts, that is a signal to add it at the account level, not just one campaign.