How to Undo Negative Keywords in Google Ads

Every account manager has the same moment. Revenue dips on one product line, nobody changed bids, nobody changed budgets, and three weeks later someone finds a negative phrase refurbished sitting in a shared list that eleven campaigns use. It took four clicks to add. It takes four clicks to remove. The expensive part was the twenty-one days in between.

The support pages that rank for this question tell you where the Remove button is. That is true and it is not enough. The interesting problem is not how to delete a negative keyword — it is finding the right one among the 4,000 in the account, proving it was the cause, and making sure the next regrettable block is reversible in minutes instead of weeks.

The exact steps to remove a negative keyword

Negative keywords live in three different places in Google Ads, and the removal path is different for each. Pick the one that matches where the block actually sits.

Ad group or campaign negatives (Search and Shopping). Select the campaign in the left panel. Open Audiences, keywords and content, then Search keywords. At the top of the table there is a toggle between Search keywords and Negative search keywords — switch to the negatives. You will see two tabs: one for negatives added directly to the campaign or ad group, one for lists applied to the campaign. Tick the row, click Remove, confirm.

Shared negative keyword lists. Click Tools, then Shared library, then Negative keyword lists. Open the list and remove the term. This changes every campaign the list is attached to, which is the point of a list and also its danger. If you only want one campaign to stop using the list, do not edit the list — go to that campaign's negative keyword table, find the list under the lists tab, and remove the association there.

Performance Max. Campaign-level negative keywords only. Open the PMax campaign, go to its settings or the negative keywords section, and remove the term. Performance Max does not accept shared negative keyword lists at all, so there is no list to detach — every negative in a PMax campaign was added to that campaign directly, which makes them easier to find and much more annoying to maintain across twelve campaigns. A PMax campaign holds up to 10,000 campaign-level negatives, and that cap matters more than most people expect once you are managing a large account.

Account-level negatives. These sit under Tools → Shared library → Account-level negative keywords on most accounts. They apply everywhere and they are the easiest ones to forget. When a term is blocked in every campaign and nobody can explain why, check here first.

Removal takes effect quickly — in practice ads can start matching that query again the same day. There is no waiting period and no approval step.

What "remove" actually does

It deletes the criterion. Google Ads does not have a disabled or paused state for a negative keyword the way it does for a positive keyword or an ad. The row disappears from the table and there is no archive you can browse later. If you want the negative back, you re-add it as a new criterion, which gets a new ID.

That is the single most important technical fact in this article, and it is why the rest of it exists.

Why the criterion ID is the thing you must keep

Every negative keyword in Google Ads has a criterion ID — a long integer that identifies that exact exclusion on that exact campaign, ad group or list. It is the only stable handle on the object. The text alone is not enough, because phrase "plumber jobs" on campaign 1234 and phrase "plumber jobs" on campaign 5678 are two different criteria with two different IDs.

You will not see the ID in the normal interface. It is exposed through the Google Ads API, in the resource name of each negative criterion, and it is what any script or tool uses when it adds or removes a block. If you work through the API, the resource name looks roughly like customers/{customer_id}/campaignCriteria/{campaign_id}~{criterion_id} — campaign and criterion, joined.

Why keep it? Because the moment you want to undo something, you need to answer three questions: which criterion, on which campaign, with which match type. The ID answers all three. The text answers none of them on its own. And once the criterion is deleted the ID is retired — re-adding the same word produces a new one, so any record that only stores the old ID points at nothing.

The practical rule is simple. Before you delete a batch of negatives, download them. Use the download button on the negative keywords table, or pull campaign_criterion / ad_group_criterion through the API. Keep the keyword text, the match type, the campaign name, the campaign ID and the criterion ID in one CSV with a date in the filename. That file is your undo stack. It costs thirty seconds and it is the difference between a five-minute rollback and an afternoon of archaeology.

Change History gets you part of the way

Google Ads does have a Change History report — Tools → Troubleshooting → Change history — and negative keyword additions and removals appear in it. Filter by change type and you can see who made the change and when. For an account with one or two people touching it, this is usually enough.

It has limits that bite in exactly the situations where you need it most. Changes made through the API or by scripts show up attributed to the connected tool or user, not to a human decision. Bulk changes can collapse into summary rows that are hard to read item by item. And the history window is finite, so a negative added two years ago by someone who has left the agency is invisible. If the term you regret was added long ago, Change History will not tell you anything. Your own CSV will.

How to find the negative keyword that is actually hurting you

Start from the symptom, not from the list. If impressions dropped on a specific ad group, the culprit is a negative that matches the queries that ad group used to win. Download every negative that applies to that campaign — ad group, campaign, every attached list, and account level — into one file, then test your known good queries against it by hand.

The nastiest cases are not exotic. Three patterns cover most of them:

A broad negative eating more than it was meant to. Someone adds broad free to stop free crm software. Broad negatives behave unpredictably across match types, and now any query containing that token is at risk — including free shipping over $50, which was converting at $40 a sale. Broad is the match type that creates regret. We dug into why in negative keyword match types don't work like positives.

A phrase negative with a word order nobody checked. Phrase plumber jobs blocks plumber jobs near me and emergency plumber jobs, which is intended. It does not block jobs plumber, which nobody searches, so it feels safe. The problem is phrase negatives like repair cost, which quietly kill boiler repair cost — a query that was the top converter in the account.

A list applied to a campaign it was never written for. This is the agency failure mode. A list called "Careers & Research" was built for a B2B software account and got attached to a local services campaign during onboarding. It blocks salary, jobs, course, training — and the local account sells training. We wrote about how to keep this from happening in managing negative keywords across client accounts.

A faster method, if you have API access or a tool: pull the search terms that stopped appearing. Compare the search terms report for a 30-day window before the suspected change against a 30-day window after. Terms that had volume and now have zero were either blocked or lost to auction dynamics. Blocked ones stop dead, not gradually. The report has its own blind spots — low-volume queries are withheld — so this works best on terms with meaningful traffic.

Six symptoms worth checking first

If you are not sure whether a negative is the cause at all, there is a shorter diagnostic: impressions down with CTR flat or up, conversion volume down with conversion rate flat, a narrow drop confined to one ad group, no bid or budget change on the date. Those four together point hard at an exclusion. We listed the full set in six signs your negative keywords are hurting performance.

The cost of being wrong, in both directions

A negative keyword you should not have added costs you the conversions you would have won. A negative keyword you were afraid to add costs you the clicks you paid for. Both are real money, and people are consistently more frightened of the first than the second.

Work the numbers on an ordinary small account. Say you spend $4,000 a month at an average cost per click of $2 — that is 2,000 clicks. In accounts we audit, the share of spend going to search terms that had no realistic chance of converting typically runs somewhere between 10% and 30%, depending on how much broad match and Performance Max is in the mix. Take the low end: 10% of $4,000 is $400 a month, $4,800 a year, gone to plumber salary and how to fix a tap yourself.

Now the other direction. You block repair cost by phrase and lose boiler repair cost, which produced 4 conversions a month at $300 each — $1,200 of revenue a month, more than the entire waste you were trying to cut. That is the asymmetry people sense intuitively, and it is why so many accounts have a negative keyword list that has not been touched in a year.

The resolution is not to block less. It is to make blocking reversible and to review it often. A block you can undo in one click is a cheap experiment. A block buried in a shared list that eleven campaigns use, added by a contractor in 2023, is a liability. If you want to put real numbers on your own account before you start, how to calculate wasted ad spend honestly walks through the arithmetic without the inflated percentages that tool vendors like to quote.

Preventing the next regret: rules that make negatives reversible

The best undo is the one you never need. Four habits do most of the work.

Default to phrase, not broad. Broad negatives are where unpredictable behaviour across match types lives. Phrase negatives still catch the variations you care about and do it in a way you can reason about. Use broad only for tokens that are unambiguous in your market — salary, vacancy, wikipedia — and never for words that appear inside your own product names.

Never block a term that converted. This is the one rule with no exceptions. If a search term produced a conversion in the last 90 days, it stays, no matter how irrelevant it looks. PPCleaner enforces this automatically — we will not block a term that converted in the last 90 days, full stop. If you do this by hand, pull the converting search terms into a column and check every new negative against it before you push.

Keep intent lists separate. One list for job seekers, one for research intent, one for competitors, one for free-seekers. Never a single list called "Negatives." When you need to undo, you want to detach a theme, not audit 900 rows. Job seeker terms in particular are safe to block in bulk for most advertisers, while competitor searches genuinely depend on the account — keeping them in different lists means you can reverse one decision without touching the other.

Date every list. Put the month in the list name or in a dummy first entry. When you inherit an account, a list named "Negatives — rebuilt 2024-03" tells you something. A list named "Negatives" tells you nothing.

What a one-click undo actually requires

For a tool to offer a genuine undo, it has to store four things at the moment of the block: the search term, the match type used, the campaign or list the negative went into, and the criterion ID returned by the API. With those four, reversal is a single delete call against a known resource name.

Without the criterion ID, a tool has to search the account for a matching text string, which is slow and ambiguous when the same word has been added in six places by six different people. That is why PPCleaner records the criterion ID for every block it makes and keeps it for 90 days. Within that window, undoing a block is one click and the ad serves on that term again that night. After 90 days the record ages out; the negative stays in place and you remove it manually, like any other.

Worth saying plainly: PPCleaner can only undo blocks it made. If a human added refurbished to a shared list in 2023, we have no record of it and no way to reverse it for you. Nothing can reconstruct history that Google never kept.

Where undoing gets harder: PMax, Shopping and the reporting gaps

Reversing a block is easy. Proving the reversal worked is harder in some campaign types than others, because the search terms data you would use to verify is not equally available everywhere.

Performance Max. You can remove the negative, but confirming that the query came back is murky. The PMax search terms report is aggregated and partial — what it actually shows is a filtered view, not the raw query list you get on Search. You remove cheap, traffic rises, and you cannot point at the exact term that returned. Give it two weeks and judge on campaign-level conversions instead. Note too that PMax has no shared lists, so a negative you want to reverse across twelve PMax campaigns is twelve separate removals. There is no bulk path short of the API or an editor upload.

Shopping. Standard Shopping takes negative keywords and shared lists like Search, so the mechanics match. What differs is the matching: you have no positive keywords, so queries are matched from the product feed, and a negative removes queries without any keyword to fall back on. Removing a negative in Shopping can reopen a far wider set of queries than you expect. Shopping campaign negatives work differently is the longer version.

Display, Video and App. These report no search terms at all in campaign_search_term_view. There are no search-term negatives to undo because there were never search terms to block. If spend is leaking there, the lever is placement exclusions, not negative keywords.

One more gap worth naming. After you remove a negative, the query does not always come straight back. The auction has moved on, your Quality Score for that query may have decayed, and competitors may now be bidding where they were not before. Expect a lag of a few days and do not panic on day two. If a week passes with nothing, the negative was not the cause — go back to the order you should check things in and work down.

A 20-minute audit of your existing negatives

If you have inherited an account and have no idea what is blocked, do this once. It is the fastest way to find the expensive mistakes.

1. Download everything. Negative keywords at account level, every shared list, every campaign, every ad group. One spreadsheet, five columns: text, match type, level, campaign, criterion ID if you can get it.
2. Sort by match type and read the broad ones first. They are a small share of most lists and they cause most of the damage. Any broad negative that is also a word in your product, service or brand name is suspect.
3. Scan for single generic words. cost, price, near, best, service, review, used. Each one made sense to somebody in a specific context. Out of that context they block high-intent queries.
4. Cross-check against your converting search terms. Export the last 90 days of search terms that produced a conversion. Any negative that would have blocked one of them comes out today.
5. Delete and log. Remove what fails the check, and save the before-and-after CSVs. You now have the undo file you did not inherit.

Twenty minutes, and you will typically find two or three blocks you would never have made. For the wider sweep, this fits inside the audit checklist you can run in an hour.

The uncomfortable part of this process is that it has no end. Queries change, the account grows, someone new joins the team. Negative keyword hygiene is maintenance, not a project. The only durable fix is to make every block cheap to reverse — record the criterion ID, keep lists thematic and dated, and never, ever block a term that converted.