If you run an established store, you already know the frustration. You open a Shopping or Performance Max campaign looking for the keyword tab, and it is not there. That is not a bug — it is the whole design. Shopping has no keywords in the Search sense, which changes how you research, target, and audit the queries you pay for.
This guide treats you as an operator with real spend and real order history, not someone launching a first campaign. The goal is to connect keyword decisions to the number that actually matters: profit per order.
Why Google Shopping has no keyword targeting
In a Search campaign you bid on keywords. In Shopping, Google decides which queries your products show for by reading your product feed in Merchant Center — your titles, descriptions, product types, GTINs, and images. Google's own guidance is explicit that complete, accurate attributes determine which auctions your products enter at all.
The practical takeaway: the title is the keyword surface. Shopping has no keyword field, so the words in your product title do the query-matching work that a keyword list does in Search. Front-load the brand, product type, and defining attributes — "Ceramic Coffee Mug 15oz Matte Black Funny Cat Design," not "Funny Mug."
This is also why so much generic advice is thin. The top-ranking articles correctly say "optimize your feed" and stop there. They skip the part you actually need as an operator: how those title words translate into spend, and whether that spend clears your margin.
The three real keyword levers in Shopping
1. Title and attribute keywords (the feed)
Because there is no bidding box, keyword research for Shopping is really research into what goes into your titles and descriptions. Practitioners report that optimized titles are associated with meaningfully more impressions and higher click-through — Store Growers' feed optimization work puts the title among the highest-leverage edits in the feed. Treat that as a practitioner pattern, not a promised outcome.
One structural caution: Merchant Center policy requires titles to match the landing page. Stuffing a title with high-volume terms the product is not actually about risks disapproval, which is the most expensive outcome of all — a disapproved bestseller silently removes the SKU carrying your account.
2. Negative keywords (the one direct lever)
Negative keywords are the only keywords you insert directly into a Shopping campaign. They do not win you queries; they stop you from paying for the wrong ones. For an operating store, the usual culprits are "free," "DIY," "template," "used," and competitor brand names that convert at a fraction of your rate.
In 2025 Google also shipped campaign-level negative keywords for Performance Max, documented in practitioner coverage of the PMax transparency update. That matters because PMax used to be a black box — now you can see and exclude the queries it buys.
3. Search-terms mining (where the real data lives)
Your search-terms report is the closest thing Shopping has to keyword research, because it shows the actual queries that triggered your products and what they did. Full search-terms reporting arrived for PMax in 2025, per the same transparency coverage, so you can now audit PMax queries the way you always could for Standard Shopping.
The workflow: pull the report, sort by spend, and split queries into three buckets. High-intent converters get worked into titles. Zero-conversion spenders become negatives. Branded queries get flagged for the cannibalization check below.
The profit angle everyone skips
Here is what no SERP competitor covers: a keyword is only worth showing for if the sale it produces clears your margin. Shopping clicks are cheap relative to Search — historical WordStream data cited by Store Growers put Shopping CPCs near two-thirds of a dollar, though that sample is from roughly 2016 to 2019 and is numerically stale today. The durable point is structural: Shopping clicks cost a fraction of Search clicks but convert at lower rates.
So the decision is never "is this a good keyword?" It is "does this query's economics clear break-even?" And break-even is pure arithmetic:
Break-even ROAS = 1 ÷ contribution margin.
Say you sell a mug for $24. Your Printify base and shipping run $15, and payment processing takes about $1. Your contribution is $24 − $15 − $1 = $8, a 33% margin. That means break-even ROAS = 1 ÷ 0.33 = 3.03x. Every dollar of ad spend on that mug has to return about three dollars just to not lose money.
Now walk a worked example at scale. Say your store does 420 orders a month at a $34 average order value, and you are weighing Google against your existing Meta spend. At a 33% contribution margin, each order contributes roughly $11.20. If a batch of generic category keywords ("coffee mug") pulls clicks at a retail-average cost — LocaliQ's 2026 benchmarks put the retail-category CPC at about four dollars and change, with a retail conversion rate near four percent — then it takes about twenty-five clicks to make one sale. Twenty-five clicks times four dollars is roughly a hundred dollars of spend for an eleven-dollar contribution. That keyword loses money at your margin, full stop.
That is the calculation the feed-optimization articles never run. A high-volume keyword that looks great in Keyword Planner can be ruinous once you put your actual margin against the actual click cost.
How keywords connect to your tROAS target
This is the operator's real control. Your Target ROAS setting is where keyword economics become a campaign instruction. Set it from the break-even math above — break-even times a profit buffer — not from a benchmark blog.
For the mug example, break-even is about 3.03x. If you want roughly a 20% profit cushion, you would set tROAS near 360%. That target tells Google's bidding to decline the auctions it predicts will not clear that bar — which is exactly how it filters out the hundred-dollar-for-eleven-dollars queries.
The trap to understand: raising the target is a volume decision disguised as an efficiency one. Google's own Target ROAS documentation warns that setting a target far above what the account has historically achieved "may limit the amount of traffic your ads may get." A target above your demonstrated capability does not create efficiency — it strangles volume, because the system simply skips auctions. The same docs note tROAS needs at least fifteen conversions with values in the past thirty days on Search and Shopping to function at all.
If you want the full break-even and bidding framework, our Google Ads economics guide walks the whole chain from margin to target.
The branded-keyword trap
The single most expensive keyword mistake for operating stores is letting automation eat your own brand. Branded queries convert cheaply because the intent is pre-built, so Performance Max gravitates to them — it chases the cheapest conversions available, "wins" sales that were coming anyway, and reports a spectacular ROAS while adding little incremental revenue.
Practitioner analyses of this cannibalization — for example from GrowthSpree — estimate that a meaningful slice of PMax budget can leak to brand queries in unprotected accounts, inflating the campaign's apparent return. Treat those figures as field estimates, not laws.
The fix exists: account-level brand exclusions for PMax, available since 2024 per AdNabu's brand-exclusion guide, plus a dedicated brand Search campaign and periodic search-term audits. To judge whether PMax is genuinely working, look at non-brand performance and total business orders against total spend — not the per-campaign ROAS number. Our breakdown of how Performance Max affects store goals goes deeper on reading those signals, and what a Performance Max campaign actually is covers the mechanics if you are newer to the format.
Feed quality beats keyword cleverness
For retailers, Performance Max spend is overwhelmingly feed-based — smec's client data, cited in their Shopping-alongside-PMax analysis, shows the large majority of PMax cost going to feed-based ads. Whatever your creative assets look like, feed quality is the load-bearing input.
That means your "keyword strategy" is mostly feed hygiene: descriptive titles, valid GTINs, correct product types, accurate availability, and price parity with the landing page. Price competitiveness in particular acts like a bid multiplier — an uncompetitive price on a shared GTIN suppresses your impression share no matter how high you bid.
Where Victor fits
Running these calculations by hand, per keyword, across a live catalog is the part that breaks down in practice. PodVector AI's Victor is an AI employee that connects to your Shopify store and Google Ads as a full operator, computes true per-order profit after product cost, fulfillment, and fees, and delivers reports to Google Drive. Victor is not a dashboard you log into to read charts — it works your live data and proposes actions, and every write action is approval-gated, so you approve before anything changes in your account. Put Victor on your store and let it do the margin-versus-click-cost math you would otherwise skip.
If you would rather hand the whole channel to a person instead, our take on Google Shopping ads agencies weighs that path.
FAQs
Can you add keywords to Google Shopping ads?
Not the way you do in Search. There is no positive keyword field — you cannot bid on a keyword to win it. Google matches your products to queries by reading your feed. The only keywords you insert directly are negative keywords, which exclude queries you do not want to pay for.
How do you do keyword research for Shopping campaigns if there are no keywords?
You research the words that go into your product titles and descriptions, and you mine your search-terms report for the queries that already trigger your products. Converting queries get worked into titles; wasteful ones become negatives. External tools like Keyword Planner give you volume context, but your own search-terms report is the decisive data because it shows what actually spent and converted.
What negative keywords should an operating store add first?
Start with intent mismatches — "free," "DIY," "template," "cheap," "used" — and competitor brand names that soak up spend without converting at your rate. Pull your search-terms report, sort by cost, and negate the highest spenders with zero conversions. Avoid negating on one slow week; wait for enough clicks to be confident the query is genuinely dead.
Why do my generic keywords cost so much and convert so little?
Because generic category terms attract browsers, not buyers, and you pay retail-average click costs for them. At a four-dollar click and a four-percent conversion rate, one sale can cost around a hundred dollars — fine on a high-margin product, ruinous on a thin-margin one. Run your break-even ROAS before deciding a keyword is worth it.
Is Performance Max replacing Standard Shopping for keyword control?
No. They now compete on equal footing in the same account, and an Optmyzr study of over twenty-four thousand campaigns found most advertisers run both, with split-budget accounts performing strongest. Standard Shopping still gives you more direct control over product groups and negatives, which is useful for forcing data on low-visibility SKUs.
How do I stop Performance Max from spending on my brand name?
Turn on account-level brand exclusions, which cover brand variants and misspellings, and run a dedicated brand Search campaign so you control those cheap, high-intent clicks deliberately. Then audit the PMax search-terms report regularly to confirm branded queries are not leaking back in. Judge PMax on non-brand results, not blended ROAS.