Keyword Planner and AI: fast research from raw list to ad groups

Google's Keyword Planner tells you three things: a term's average monthly search volume, its competition level and an estimated top-of-page bid range. From one seed term it generates hundreds of ideas and lets you export them.

What it does not give you is judgement. It will not say which term carries purchase intent, which one is looking for information, or which ones belong in the same ad group. Grouping a five-hundred-row list by hand is the longest part of the research.

With the Tecrube connector, telling Claude or ChatGPT “pull keyword ideas for orthopaedic mattress with volume, competition and bid range, then group by intent” returns the list as a draft of ad groups rather than a raw table: purchase-intent terms, comparison terms, information seekers and brand terms, each separate.

The information-seeking group is usually the first draft of the negative list: patterns such as “what is”, “how to choose” or “reviews”. Research and the negative list come out of the same step instead of being two jobs.

Once the groups are set, the match type decision follows. Phrase match is a safe start for purchase-intent terms with known volume; broad match only makes sense with Smart Bidding and solid conversion data; exact match is reserved for brand and the few most valuable terms.

Ask the assistant for a match type suggestion per group with its reasoning, and make the call yourself. Opening with phrase match and widening to broad as the search terms report fills up is cheaper than the reverse.

Document the match type decision: noting in the campaign notes why each group opened on phrase or broad saves you time when interpreting performance three months later. You can ask the assistant to draft that note alongside the build preview.

Planner volumes are ranges and averages; seasonality and regional differences are hidden. “Air conditioning service”, for example, looks reasonable on an annual average while demand clusters in summer. Asking the assistant for the monthly spread and planning the budget around it saves accounts that averages mislead.

The competition indicator is the number of advertisers, not organic competition; low-competition terms with volume can be the cheapest conversions in the account. They are worth tracking in a group of their own.

When the grouping is done, the build continues in the same chat: “create a search campaign with these groups; add two RSAs, sitelink and callout extensions per group, put the information terms in the negative list, $40 daily budget”. Tecrube presents the whole campaign in one preview; you approve and it is created paused.

Research is a read and spends no credits; the build is a write and follows preview → approve → apply. The 75 free credits cover one research pass and one campaign build; Pro at $49 a month includes 5,000 credits.

The Planner can generate ideas not only from a seed keyword but also from a web address. Giving a competitor's category page as the source surfaces product names and user phrasing you had not thought of. Ask the assistant to compare that list with your own seed list and show only what is new.

The most valuable source of ideas, however, is your own account. Search terms in the report that convert but have not been added as keywords are a list of proven demand. In Tecrube, “show the converting search terms that are not in my keyword list, with Planner volumes” combines both sources in one table.

If Search Console is connected, a third source joins: queries that earn organic impressions but that you never bid on. A list built from all three is far more accurate than research started from scratch.

You can narrow the combined list by telling the assistant “score these by purchase intent, estimated cost and our current organic position, and pick the top twenty”. That way the research ends not with a table of hundreds of rows but with a short list ready to go into a campaign. The whole step is a read; it spends no credits and can be repeated as often as you like.