How to read the Google Ads search terms report

There are about a dozen columns visible in the search terms report, and only three of them matter for the first sweep. Knowing which three saves twenty minutes of the wrong kind of attention.

The report is dense by design. Google built it for power users, and the default column set reflects that. Once you know which columns are doing the load-bearing work, the rest become reference material you only consult when the first three leave a question unanswered.

The three columns that matter

The work the search terms report exists to support is finding queries that ate spend without paying back. Three columns answer that question.

Search term. The actual query a user typed. This is what you read to make the keep-or-block decision.

Cost. What that query cost over the date range you set. Sort by this column descending and the highest-cost rows rise to the top.

Conversions. Whether the query led to a measured conversion. A high-cost row with zero conversions and bad intent is a negative; a high-cost row with several conversions is a keep.

Three columns. Everything else is decoration on the first sweep.

What the other columns are for

The rest of the columns answer secondary questions that come up only after you have already triaged the high-cost rows.

Match type. Tells you whether the matcher classified the match as exact, phrase, or broad-match-and-close-variants. Useful for diagnosing why a query slipped through despite your match-type intentions, but rarely the answer to the first-pass keep-or-block decision.

Added/Excluded. Shows whether you have already added the term as a keyword or as a negative. The "None" value, which is most rows, means the term came in via match-type expansion and is up for review. The "Added" value means you already bid on it explicitly. The "Excluded" value means a negative is already in place; if a row says Excluded and is still costing money, your negative is not catching what you thought.

Impressions. How many times the ad was shown for that query. Useful for high-impression-low-click rows that suggest a relevance or copy problem, but those are diagnostic patterns for a later sweep, not for the first pass.

Clicks. How many people clicked the ad after seeing it for that query. Useful as a sanity check on the cost column; if cost is high but clicks are low, your CPC for that query was high. That is a separate diagnostic question.

CTR. Click-through rate (clicks divided by impressions). Almost never the answer to the first-pass question. A low CTR for a junk query is fine; a low CTR for a buyer query is a copy problem.

Avg. CPC. What you paid per click on that query. Useful for diagnosing why specific high-spend rows ate so much of the budget; sometimes the answer is high CPC on relatively few clicks.

Conversion rate, Cost / conv., Conv. value. All useful for evaluating performance of accepted (kept) queries. Not useful for the keep-or-block decision, which is binary and intent-based.

Reading order, top to bottom

Good readers of the report do roughly the same thing in the same order. The order matters; switching columns mid-row breaks the rhythm and burns time.

For each row:

1. Read the search term. Form a quick mental picture of who typed this and what they wanted.
2. Glance at cost. Is this row big enough to matter, or is the spend negligible?
3. Glance at conversions. Did the query pay back?
4. Decide: keep, negative, or come back later.
5. Move to the next row.

That cycle, repeated fifty times, takes about fifteen minutes if you stay in the rhythm. It takes forty-five minutes if you let yourself check match type, impressions, and CTR on every row. The slower path produces marginally better decisions, but the marginal improvement does not outweigh the time cost on the first pass.

The exception: a small number of rows will be ambiguous on the first three columns. The query is borderline, the cost is meaningful, and the conversions are zero. For those rows, glance at impressions and clicks. Sometimes a row with high impressions and few clicks is genuinely a poor-fit query; sometimes it is a copy problem. The distinction often clarifies the decision.

What the search-term column is actually showing

The search term is not always the literal query the user typed. Google sometimes shows a slight normalization: lowercased, de-duplicated whitespace, and occasionally normalized punctuation. For most rows this is invisible.

There are two cases where it matters.

The "(other)" rows. Some rows show up as "(other search terms)" and group together queries that fell below the privacy threshold. You cannot see what those queries were. The cost and conversion totals are still attributed correctly, but the term-level detail is hidden.

Misspellings and variants. When the matcher catches a misspelling, the report sometimes shows the user's misspelled version and sometimes shows the corrected version. This is more common in 2026 than it used to be because of close-variant expansion. The takeaway: if you see a phrase that looks slightly off, do not assume it is a typo in the report. The user probably typed it that way.

What "added/excluded" tells you that is non-obvious

The Added/Excluded column has three values, and one of them is informational and two are actionable.

None (informational). The query came in via match-type expansion; you did not explicitly bid on it. Most rows are this value. None just means "this is a query the matcher decided to serve against on its own, without a specific keyword from you."

Added (actionable, sometimes). You have this term as an explicit keyword somewhere. Useful to know; if a query is performing well and you have it as a keyword, it is doing what you wanted. If a query is performing badly and you have it as a keyword, the problem is on the keyword side (bid or copy) rather than a negative-keyword question.

Excluded (actionable, almost always). A negative is in place. If a row says Excluded and the cost is non-zero in the date range, something is wrong. Either the negative was added recently and the row's cost predates it, or the negative is not catching what you thought. Click into the row to see when the cost was incurred.

The Excluded-with-cost case is rare but worth catching. We have seen accounts where a negative was added at the wrong level (campaign instead of shared list, or ad group instead of campaign), and the report kept showing the term as Excluded while the matcher kept serving against it through a different campaign that did not have the block.

Filters that pay back

Five filters get used regularly in disciplined sweeps. The first two are universal; the rest are situational.

Clicks above one. Cuts the row count by roughly a third without losing any meaningful signal. A query with one click and zero conversions is statistically almost meaningless.

Cost above ten dollars (or whatever threshold matches your account size). For sweeps, focus on rows that ate at least ten dollars of spend. Below that, the leakage per row is small enough to leave for the long-tail sweep.

Match type = Broad match. Useful when you want to specifically audit broad-match expansion. Filtering to broad rows surfaces the queries the matcher pulled in beyond your literal keywords, which is often where the most surprising junk hides.

Conversions = 0. Combined with cost > threshold, this is the fastest filter for "where did the wasted money go." Apply it once, scan, then remove it before reviewing the keep candidates.

Date range = last 30 days. Always. The seven-day default does not have enough volume on most accounts. The ninety-day range dilutes recent shifts. Thirty days is the sweet spot.

Exporting versus reading in the UI

The report is fine to read inside the Google Ads UI for sweeps under fifty rows. Beyond that, exporting to CSV is faster.

The export gives you all visible columns plus a few extras (campaign, ad group, date range used). The CSV opens in Excel, Numbers, or Google Sheets cleanly. From there, sorting, filtering, and tagging are faster than the same operations inside Google Ads.

The other reason to export: you can paste the column of negative-flagged terms directly into a separate CSV that becomes your bulk-upload source. The whole audit, from open-the-report to apply-the-negatives, can live in one spreadsheet.

We built Sensei Ads so that the export-and-tag step happens automatically: the CSV goes in, the report comes back categorized with reasons attached, and the tagged list drops back into Editor. The same workflow is doable manually; the tool just shortens the per-row decision time.

What to do with the report after the sweep

The report is a working surface. Once you have applied the negatives from a sweep, the report's job is done until next month. Do not save it as a record, do not screenshot it for archival, do not build dashboards around it. The journal of what you blocked, with reasons, is the durable artifact. The report just regenerates itself the next time you need it.

Two specific don'ts:

Do not build "trended search terms reports" that compare month over month. The report changes faster than the trend would imply. The privacy threshold shifts, match types tweak, query distribution moves. Trending it is mostly noise.

Do not show the raw report to clients. It is dense, technical, and easy to misread. The artifact for clients is the summary of what you blocked and why, and the resulting impact on cost-per-acquisition. The report itself is internal.

The sweep ends when the negatives are applied. Move on.