Comparison queries are the most misjudged bucket in the search terms report. They do not read as waste, because the subject matches your product. They do not read as a clean win either, because half of them come from people building a shortlist who will not buy from an ad today. A query like best project management software, asana alternatives, or monday vs asana lands in your report on broad and phrase match, spends money, and gives you no label telling you whether it was a shopper two clicks from converting or a researcher three weeks out. Treat the whole bucket as waste and you cut off some of your highest-intent demand; keep all of it and you fund a steady stream of list-builders. The job is to split it, and the query wording gives you enough to do that without guessing.
This is a different decision from the two that look adjacent. Negating a competitor brand name turns on whether you have a campaign built to compete for that one rival; cutting an almost-relevant term turns on the query having the wrong intent entirely. Comparison terms are neither. They often carry the right intent at a specific funnel stage, and a single query can be a buyer or a browser depending on one word in it. So the rule here is not "is this relevant" or "is this a competitor" — it is "is my product actually in this comparison, and can the page I send the click to win it?"
What counts as a comparison-intent search term
A comparison-intent term is any query whose structure pits options against one another instead of naming a single thing to buy. The reliable tell is a comparison modifier, and they fall into a few families: superlatives (best, top, cheapest, fastest), the alternatives family (alternatives, alternative to, similar to, like X), the head-to-head family (vs, versus, compared to, or as in asana or monday), and the evaluation family (review, reviews, comparison, ranked). When one of these words sits next to your product, category, or a competitor, you are looking at a comparison term, not a plain product query.
The reason this bucket deserves its own treatment is that the modifier changes the intent more than the noun does. Project management software and best project management softwarecontain the same product, but the second is a person actively weighing choices, while the first could be anyone from a buyer to a student writing an essay. That shift is why you cannot fold comparison terms into your general irrelevance scan: the noun passes the relevance test every time, so an irrelevance pass waves them through, and a zero-conversion pass cannot tell a term that has not converted yet from one that never will. The modifier is the signal, and reading it deliberately is the whole technique.
Why they appear when you never added those keywords
You see comparison terms because broad and phrase match expand your keywords into the queries people actually type, and comparison modifiers are among the most common expansions. A broad keyword like crm software can match best crm software, crm software alternatives, or crm software vs spreadsheet without you ever adding a single comparison word, because close variants and broad matching treat those as the same underlying intent. The report is where the expansion becomes visible; the keyword list is not, because the keyword you added looks innocent. This is the normal mechanism of a broad-heavy account, not a malfunction — see how the search terms report surfaces what your keywords matched for the full picture.
It also sets a hard constraint on your fix. Negative keywords only block the exact terms you list, and crucially they do not match close variants, so a negative on alternatives will not stop alternative or a misspelling, and a negative on vs will not catch versus. If you decide a modifier family is waste for a campaign, you have to add each form you want gone — singular, plural, abbreviation, synonym — as explained in how negative match types handle plurals and close variants. That is more work than it looks, which is another reason to negate only the modifier families you are certain about and leave the mixed ones to a per-query decision.
The one question that decides it: are you in the comparison?
For any comparison term, the deciding question is whether your product is genuinely one of the options being weighed, and whether the page you would send the click to can win that weighing. If both are true, the query is demand you want. If either is false, it is a researcher cost. This single test cuts through the bucket faster than any list of good and bad modifiers, because the same modifier can go either way: best email tool is gold if you sell an email tool with a comparison-ready landing page, and pure waste if you sell email templates and the click bounces off a mismatched page.
Apply it by reading the full query, not the modifier alone. Your-category alternatives and competitor vs your-brand put you squarely in the comparison — keep them. Competitor-A vs competitor-B, where neither is you, is a comparison you are not part of — negate it, because the best you can do is interrupt a decision between two other products. Best [thing you do not sell]is a shopper for someone else's shelf — negate it. The query text almost always tells you which case you are in; the cost and conversion columns then tell you how urgent the call is.
Keep: switch-ready and decision-stage comparisons
The comparison terms worth protecting are the ones where someone has narrowed to a short list and is looking for a reason to pick. Superlative queries are the clearest case: Neil Patel's team argues that searchers using best or top are looking for help choosing and should be mapped to bottom-funnel offers rather than dismissed. When your product is a legitimate answer to best [your category], that click is a decision-stage shopper, and negating it hands the decision to whoever did keep the term. The same holds for [competitor] alternatives when you are a credible alternative — the searcher has already decided they want to switch and is auditioning replacements.
Keeping these terms is only half the move; they have to land somewhere that wins the comparison. A switch-ready alternatives query sent to a generic homepage converts far worse than one sent to a page that actually addresses the comparison, so the protect decision implies a destination, not just an unblocked keyword. Where volume justifies it, pull the high-intent comparison queries into their own ad group or campaign with ad copy and a landing page built for the comparison, the same structural logic you would use to deliberately compete for a rival's brand term. The point is that these queries reward intent-matched handling and punish lazy handling — so the risk of keeping them badly is real, which is exactly why so many accounts over-negate them instead of doing the work.
Negate: researcher and out-of-set comparisons
The comparison terms to cut are the ones where you are not in the set, or where the modifier reveals a researcher rather than a shopper. Competitor-A vs competitor-B with no mention of you is the cleanest negate: you are paying to appear in a decision between two products that are not yours, and the realistic outcome is an expensive, confused click. Best [segment or use case you do not serve] is the same problem in superlative form — the subject matches your category loosely enough to trigger the ad, but the specific thing being compared is not something you sell. These are close cousins of the almost-relevant terms that read as on-topic while carrying an intent you cannot convert.
There is also a volume-and-economics layer. Some comparison terms are genuinely mixed and low-volume — one click here, one there, never enough to judge a single row. Do not try to negate those individually; a one-off obscure-tool vs other-obscure-tool query will not recur in that exact form, so a negative on it buys nothing. Instead, negate the wasteful pattern when one is clear — an entire family of comparisons against products you do not compete with — or let Smart Bidding bid the non-converting ones down on its own. Reserve manual comparison negatives for the terms that both recur and clearly sit outside your set, and resist the urge to prune every ambiguous row, because over-negating throttles the good comparison traffic along with the bad.
How to action it in the report this week
Open the search terms report over a 60 to 90 day window so the comparison tail has time to show up, then filter the term text to surface the bucket: filter for terms containing best, top, alternative, vs, versus, compare, and review, running the filter once per modifier since one filter cannot catch all forms at once. Add the Cost and Conversions columns and sort by cost descending. You now have your comparison bucket ranked by how much each term is costing you, which is the order to make decisions in — the expensive, recurring terms are where a right or wrong call actually moves money.
Go down the list and apply the one question to each: are you in this comparison, and can your page win it? Keep the switch-ready terms and note any that deserve their own ad group or a comparison landing page. Negate the out-of-set and researcher terms, adding each modifier form you mean to block rather than assuming one covers the variants. Leave the low-volume ambiguous rows alone. Do this at the right level — a comparison that is universally irrelevant to a campaign belongs as a campaign-level negative, while one that should be blocked in a generic ad group but kept in a comparison-specific one belongs at the ad-group level. The whole pass takes under an hour on most accounts, and the output is a cleaner comparison bucket plus a short list of high-intent terms you were about to throw away.
The bottom line
Comparison search terms are not a relevance problem and not a competitor problem — they are an intent-stage problem, and the stage is often exactly the one you most want to buy. The mistake that costs the most is the reflex to treat every best, alternatives, and vs query as noise and negate the lot, because buried in that bucket is some of the most decision-ready demand in the account. The opposite mistake, keeping all of it, funds list-builders and out-of-set comparisons that will never convert through an ad.
The one question splits the bucket cleanly: is your product actually in this comparison, and can the page you send the click to win it? Keep the terms where both are true, and give the valuable ones a destination that matches their intent. Negate the ones where you are not in the set or the modifier reveals a researcher, adding each variant form because negatives do not catch close variants on their own. Leave the low-volume ambiguous rows to pattern negation or the bidder. Done once a quarter on top of your normal search-terms review, this turns a confusing column of comparison queries into a deliberate call on each one — and stops you from negating your way out of the demand you were trying to capture.