Shopping Campaign Negative Keywords Work Differently

Shopping matches like Performance Max and takes negatives like Search

Standard Shopping has no positive keywords. Google decides which queries your products can answer by reading the feed — title, description, product type, Google product category, brand, attributes. That half of the campaign behaves like Performance Max: you supply products, Google supplies the matching.

The other half behaves like a Search campaign. Standard Shopping accepts negative keywords at campaign level and at ad group level. It accepts shared negative keyword lists. It supports broad, phrase and exact negatives. You can write all of it through the UI or the API. Performance Max accepts campaign-level negative keywords only, capped at 10,000 per campaign, and will not take a shared list at all.

Almost every guide to Shopping negatives treats the campaign as "Search without keywords" and hands you a list of 1,500 generic blocks — jobs, salary, free, torrent, craigslist. That list is not wrong. It is just aimed at the wrong problem. The generic waste list was built for Search campaigns whose broad-match keywords wandered. Shopping doesn't wander; it answers whatever the feed looks like it answers. The waste is shaped by your product titles, and no one else's list knows what your product titles say.

In Shopping, negatives are not a trim — they are the shape of the campaign

Because there are no positive keywords, the negative list is the only keyword-level definition of who your campaign is for.

Take a Shopping campaign selling Bosch dishwasher spare parts. Feed titles read like "Bosch Dishwasher Door Seal — SMS46KI01G". Over a month the search terms report fills up with:

• bosch dishwasher door seal
• bosch dishwasher door seal replacement
• bosch dishwasher door won't close
• bosch dishwasher e24 error
• bosch dishwasher manual pdf
• dishwasher repair near me
• dishwasher engineer jobs
• how to remove a dishwasher door
• used bosch dishwasher gumtree

One of those is a person about to buy a door seal. Two are ambiguous. The rest are a repair trade, a support library and a second-hand marketplace, all of them crowding into your campaign because your product title contains the words "Bosch", "dishwasher" and "door".

You did not opt into any of them, and you cannot un-opt by editing a keyword, because there is no keyword. You either block them or you pay for them.

This changes what a good Shopping negative list looks like. A Search negative list tends to be a record of mistakes: the odd term a broad keyword caught. A Shopping negative list is a systematic map of the non-transactional halo around your product nouns — the repair intent, the job intent, the manual intent, the rental intent, the marketplace intent. It grows faster and it grows in a more predictable direction, which is exactly why it responds well to being maintained nightly rather than reviewed monthly. If you are worried the list is getting long, the length itself is rarely the problem — see how many negative keywords are too many for where the real limits bite.

Three things Shopping can do that Performance Max cannot

Shared lists, ad group negatives and priority tiers. Each one is a genuine structural advantage, and each one has a catch.

Shared negative keyword lists

Standard Shopping takes shared lists, like Search. You build one "never a customer" list — jobs, salary, DIY, repair manual, free download — attach it to every Shopping campaign, and maintain it in one place. Add a term once, it applies everywhere.

Performance Max does not accept shared lists. Campaign-level negatives only, and a separate 10,000-keyword bucket per campaign. If you run Shopping and PMax side by side, the same blocked term has to be written once to a shared list and then again into each PMax campaign individually. That asymmetry is the single biggest operational difference between the two, and it is covered in detail in negative keywords in Performance Max and the 10,000 cap.

Ad group level negatives

Shopping lets you put negatives on an ad group. This is how you split branded traffic from generic traffic without duplicating campaigns. The same product can sit in two ad groups; the negatives decide which ad group is allowed to answer which query.

A common setup: a "brand" ad group and a "generic" ad group over the same products. The generic ad group carries bosch, neff, siemens as phrase negatives so it never serves on brand searches. The brand ad group carries negatives for everything except the brand terms. Now you can bid differently on the two intents with one campaign and one feed.

The catch: ad group negatives are invisible from the campaign view and easy to forget. Six months later someone wonders why a product with plenty of impressions never shows on its own brand name. The answer is three clicks deep in an ad group you did not open.

Priority tiers

The campaign priority setting — low, medium, high — decides which of your Shopping campaigns enters the auction when more than one of them contains the same product. Higher priority wins, regardless of bid. Negatives are what push a query down from one tier to the next.

The classic build: a high-priority catch-all campaign with a low bid takes everything, with negatives for your known converting terms. Those terms fall through to a medium or low-priority campaign with a higher bid. You end up paying more for the queries that earn and less for the rest, from one feed.

Say the hard part. In a tiered structure a negative does two jobs at once: it stops waste and it reroutes traffic. Add a term to the wrong tier and you have not saved a penny — you have just moved the spend to a campaign with a higher bid. This is the one situation where automated negative keyword management needs a human to have decided the routing logic first. PPCleaner writes negatives to the campaign that spent the money; it does not know that your high-priority campaign is a funnel rather than a destination.

Match types behave the same, but the blast radius is bigger

Broad, phrase and exact work in Shopping exactly as they do in Search. What changes is how much damage a sloppy broad negative does, because there is no positive keyword holding a floor under your traffic.

In a Search campaign, a bad broad negative blocks some of the queries your keyword would have caught. In Shopping, everything the feed can match is in play, so a bad broad negative can remove an entire commercial intent you never realised was there.

The example that catches people: a broad negative free, added to stop "free dishwasher" and "free delivery" queries, also blocks free standing dishwasher when the searcher types it as two words. A broad negative part to stop "spare parts wanted" also blocks dishwasher door part number. A broad negative used blocks used to work now leaking — a genuine repair-part buyer describing their problem.

Two rules that keep this manageable:

1. Default to phrase. dishwasher jobs as phrase blocks the job hunters and leaves jobs lane appliance repair alone. Broad negatives are worth it only for words that can never appear in a buying query in your category — and there are fewer of those than you think.
2. Add the variants yourself. Google documents that negative keywords do not match close variants. dishwasher will not block dishwashers, dish washer or dishwaher. Your positive-side intuition about plurals and typos does not transfer. The full asymmetry is laid out in negative keyword match types don't work like positives.

What the Shopping search terms report shows you, and what it hides

Search terms for Search, Shopping and Performance Max all report through campaign_search_term_view. Display, Video and App report none at all — if you are running those, there is no query data to act on, and no negative keyword to write.

For Shopping you get the query, the campaign and the usual metrics. What you do not get, cleanly, is the product that answered the query. The search terms view is not a per-product view, so the link between "this query cost $14" and "this SKU's title dragged it in" is something you infer from the words, not something the report hands you. In a 50-product feed that is fine. In a 40,000-product feed it means you are treating symptoms.

You also do not get every query. Google withholds search terms that do not meet its privacy threshold, so a chunk of your Shopping spend sits under no query at all. It is real money and it is unblockable — no term, no negative. The search terms report and what it hides goes through what that missing block is actually made of and how to size it.

One consequence worth stating plainly: you cannot negative your way to a clean Shopping campaign, because you cannot see all of it. Negatives handle the visible waste. The invisible portion is handled by the feed, by product exclusions and by bidding.

Terms worth blocking — and the ones that look blockable but convert

The blockable categories in a Shopping account are boringly consistent. Working from the dishwasher-parts example, they are:

• Employment. dishwasher jobs, kitchen porter salary, appliance engineer apprenticeship.
• Service and repair intent — if you sell parts but not labour. dishwasher repair near me, appliance engineer call out, who fixes dishwashers.
• Documentation. bosch dishwasher manual pdf, smv46 wiring diagram, e24 error meaning.
• Second-hand marketplaces. gumtree, ebay, facebook marketplace, preloved.
• Rental and hire. dishwasher rental, appliance hire.
• Adjacent products you do not stock. dishwasher tablets, dishwasher salt — if your feed is parts only, these are somebody else's revenue.

And then the ones that look like waste and are not. These are the terms an over-eager negative list kills, and they are the reason a block list should never be copied wholesale from a blog post:

• cheap. Someone searching cheap dishwasher door seal is shopping for a $25 part. Price-sensitive is not the same as non-buying.
• near me. If you run local inventory ads, these are your best queries.
• how to fix a leaking dishwasher. Genuinely ambiguous. If you sell the seal that fixes it, some of those searchers buy the seal. Look at the data before assuming.
• Competitor retailer names. Sometimes terrible, sometimes your cheapest conversions. Category-dependent, account-dependent.
• Model numbers and part codes. They look like gibberish in a report. They are frequently the highest-intent queries in the whole account.
• Misspellings of your own brand. Easy to sweep up in a typo-blocking pass, expensive to lose.

This is why PPCleaner never blocks a term that converted in the last 90 days, whatever it looks like. A term that has taken money is not a candidate, full stop. If you are doing this by hand, apply the same rule before you apply anyone's list. How to find wasted ad spend in Google Ads sets out a way to sort the report that puts the real candidates at the top.

Negatives cannot fix a bad feed

If your product title is "Waterproof Jacket — Blue — Large", you will be matched to dog coats, car covers, children's sizes and motorcycle gear. You can block those queries one by one, forever, as the language shifts. Or you can change the title to "Men's Waterproof Rain Jacket, Blue, Large" and watch most of them stop arriving.

Negatives are the brake. Feed titles are the steering. A negative list that keeps growing in one direction is not telling you to add more negatives — it is telling you a product title is ambiguous.

PPCleaner does not touch your feed, your titles, your budgets, your bids or your product groups. It reads what you paid for and blocks what was never going to convert. That is a real limit. If you notice that eighty of your last hundred blocked terms all cluster around one misleading word in one product type, the fix is upstream of anything a negative keyword tool can do.

The same goes for the other things negatives cannot repair: missing GTINs, wrong Google product categories, uncompetitive pricing, disapproved products. None of those show up as a bad search term. They show up as no search terms at all.

Running Standard Shopping next to Performance Max

Google documents that Performance Max takes precedence over Standard Shopping for the same products in the same account. The Shopping priority setting does not apply to PMax; PMax simply wins.

That has a direct consequence for negatives. If you add dishwasher repair near me to your Shopping campaign and nothing changes, it may be because PMax was serving that query all along. The negative worked perfectly. It just protected a campaign that was not spending the money.

So when you run both:

1. Check campaign_search_term_view for which campaign the query landed in before deciding where to write the negative.
2. Assume any block that lives only in Shopping will relocate the query rather than stop it, as long as PMax carries the same products.
3. Remember that the Shopping side can hold the block in a shared list and the PMax side cannot. Each PMax campaign needs its own copy, inside its own 10,000-keyword budget.

The cleanest arrangement is to treat the two as one blocklist maintained in two formats: a shared list attached to every Shopping campaign, and a mirrored set of campaign-level negatives in each PMax campaign. That is tedious by hand and trivial to automate, which is roughly the argument for automating it.

What a nightly pass actually does

Every night, PPCleaner reads the search terms your account paid for across Search, Shopping and Performance Max. It judges which ones were never going to convert, and writes them as negative keywords — to the shared list or campaign for Shopping, to the campaign for PMax, in the match type that blocks the term without collateral damage. Anything that converted in the last 90 days is skipped. Every block is listed, and every block can be undone in one click for 90 days.

It does not change budgets, bids, ads, keywords, product groups or the feed. It does not know your margin on a specific SKU, and it does not know that you are deliberately buying an unprofitable term this quarter for reasons that are not in the data. Read the log. When it blocks something you wanted, undo it — and that term stays allowed.