Dynamic Search Ads Negative Keywords That Actually Hold

Why DSA search terms drift away from your pages

Dynamic Search Ads drift because there is no keyword list holding them in place. Google crawls your site, decides what each page is about, and matches searches to the page it thinks fits best. Every page it can reach is a possible entry point, including the ones you never meant to advertise.

In a normal Search campaign, you write "emergency plumber" and Google matches searches to that keyword, close variants included. In a DSA campaign the "keyword" is your site's text. If your plumbing company has a blog post called "How to fix a leaking tap in 10 minutes", Google now has a good reason to show your ad for "how to fix leaking tap yourself". That page really is relevant to that search. It just won't bring you a customer.

The same thing happens across almost every site we audit:

• Blog and guide content brings in informational searches: "how to", "what is", "diy", "symptoms of".
• Careers pages bring in job seekers: "plumber jobs", "electrician apprenticeship", "hvac technician salary".
• Support and login pages bring in existing customers looking for help: "acme software login", "cancel acme subscription".
• Out-of-stock or discontinued product pages keep matching searches for products you can't sell.
• Location pages for areas you no longer serve keep pulling local searches from those areas.

Then there is a second layer of drift on top of the first. Google doesn't only match the exact words on the page. It matches searches it thinks are related to the page's topic. So a page about "commercial boiler installation" can serve for "boiler repair cost", "boiler not working" or "boiler engineer course". The page didn't change. Google's reading of it got broader.

This is why DSA search terms reports look messier than keyword-based ones. You didn't choose most of the queries. You chose a set of pages, often without realizing it, and Google chose the queries.

Three controls, three different jobs

DSA gives you three ways to control what happens, and each one works at a different level. Page feeds decide which pages can serve. Exclusions remove pages or page types. Negative keywords remove searches. Most guides treat these as separate topics. In practice they only work well when used together.

Page feeds: decide which pages can serve

A page feed is a list of URLs you upload, optionally with custom labels. You can then target "all URLs in this feed" or "URLs with the label services", and set the campaign to use only URLs from your page feed. Once that setting is on, Google stops pulling landing pages from the rest of your site.

This is the strongest control you have, because it removes whole categories of drift at once. If the blog isn't in the feed, blog posts can't be landing pages, and most of the "how to" traffic goes with them.

The cost: you lose some of what makes DSA useful. Part of the point of DSA is to find searches and pages you hadn't thought of. A tight page feed turns DSA into something closer to a keyword-less, page-based Search campaign. For many advertisers that is exactly right. For a large e-commerce catalogue with constant new products, keeping the feed up to date becomes a job in itself.

URL and content exclusions: remove pages

Dynamic ad target exclusions let you tell Google which pages must never serve. You can exclude by URL contents (/blog/, /careers/, /support/), by page title, or by page content (for example, pages containing "out of stock" or "no longer available").

Exclusions suit advertisers who want to keep broad targeting, such as "all webpages" or "categories", and only cut off the parts of the site that obviously won't convert. They are quicker to set up than a full page feed, and they keep DSA's discovery side.

The hard part: exclusions depend on your site structure. If your blog posts live at /how-to-fix-a-leaking-tap with no /blog/ folder, there is no clean URL rule to exclude them. Page-title rules help, but they are fragile. One rewritten title and the page is eligible again. Exclusions also do nothing about the second layer of drift, where a good page serves for a bad query.

Negative keywords: remove queries

Negative keywords work in DSA the same way they do in any Search campaign. They are matched against the search, not against your page. You can add them at the ad group or campaign level, and you can apply shared negative keyword lists to DSA campaigns just as you would to standard Search campaigns.

Negatives are the only control that handles "good page, wrong intent". Your boiler installation page is a page you want to advertise. "Boiler engineer course" is still a search you never want to pay for. No page feed or exclusion can separate those two. A negative can.

The three match types behave as they always do. Broad, phrase and exact are all available. Broad negatives behave unpredictably across match types, so for DSA, where the query set is already wide, phrase is usually the safer default for intent words ("jobs", "course", "how to") and exact for specific searches you want gone without side effects.

How to tell which control a bad term needs

The quickest test is to look at the landing page Google picked for the search term. If that page should never be a landing page, fix the page targeting. If the page is fine and the search is wrong, add a negative. If you see the same pattern across many pages, tighten the feed.

The DSA search terms view shows the search term next to the landing page that served. That second column is what makes DSA diagnosis different from a normal search terms review. In a keyword campaign you ask "which keyword matched this?". In DSA you ask "which page matched this, and should that page be here at all?".

Two reasons this order matters.

First, a negative keyword added for a page problem never stops growing. Block "plumber jobs" and the careers page shows up next for "plumbing apprenticeship", then "trainee plumber", then "plumbing vacancies". Each one is a new negative. Excluding /careers/ ends the whole family in one change.

Second, an exclusion added for a query problem cuts off good traffic. If you exclude the boiler installation page because it attracted "boiler engineer course", you also lose "boiler installation quote", which was the reason the page was there.

Not every case is clean. Sometimes the page is half right. A pricing guide like "How much does a new boiler cost?" brings both buyers and researchers. In that case keep the page and use negatives for the clearly informational versions ("average", "calculator", "uk", if you only serve the US). Then judge the rest on conversion data over a longer window rather than on the wording of the search.

Setting up negatives for a DSA campaign

Start with a shared list for the intent words that are wrong for your business anywhere, then add campaign-level negatives for problems that are specific to DSA. Keep the two separate so you can tell later why a term was blocked.

A shared list for universal intent

Most accounts already have a list of words that never convert: "jobs", "salary", "free", "diy", "course", "training", competitor names you choose not to bid on. Apply that list to the DSA campaign on day one. DSA finds these searches faster than keyword campaigns because it has more pages to match them to. The trade-offs of account-level versus campaign-level lists apply here exactly as they do elsewhere.

Be careful with words that can mean buying intent in some contexts. "Cheap" and "free" are the classic examples, and they need more thought than a blanket block. We cover that in blocking "free" and "cheap" searches without losing buyers.

Campaign-level negatives for DSA drift

These are the terms that only show up because of how DSA reads your pages: topic drift ("boiler not working" when you only install), adjacent services you don't offer, and brand searches for tools or products you mention in your content but don't sell. A page comparing three accounting tools will pull searches for all three brands.

Overlap with your keyword campaigns

Many advertisers run DSA next to standard Search campaigns, using DSA to catch searches the keyword list misses. If you want DSA to stay in that role, you can add your main exact keywords as exact-match negatives in the DSA campaign. That pushes those searches back to the campaigns where you control the ad text and bids.

The cost of doing this: you give up DSA as a fallback when a keyword campaign is limited by budget or paused. And if a negative in DSA collides with a keyword you meant to keep live elsewhere, nothing in the interface warns you. These are the negative keyword conflicts Google Ads won't flag, and DSA campaigns are where they tend to pile up.

Don't judge only on the obvious terms

The big, obviously wrong terms are easy. The harder part of a DSA search terms report is the long tail: hundreds of searches with one or two clicks each, none worth a decision on its own. DSA produces more of these than keyword campaigns because its matching is wider. Together they can be a real share of spend, which is why low-volume terms add up is a bigger issue here than almost anywhere else. Some of the long tail will also be hidden from you entirely. Google doesn't report every search, and the search terms hidden by privacy can't be negated one by one.

A worked example

Here is how the three controls split the work in practice. The numbers are illustrative, but the pattern is one we see often in accounts we audit.

A home services company runs one DSA campaign targeting "all webpages". It spends $4,000 a month. Its site has service pages, area pages, a blog with about 80 posts, a careers page and a customer login.

Looking at a month of DSA search terms grouped by landing page:

• Service and area pages: $2,800
• Blog posts: $700
• Careers page: $300
• Customer login and support: $200

That adds up to $4,000.

Step 1, exclusions. The careers page and the login and support pages should never be landing pages. Excluding /careers/, /login and /support/ removes $500 of spend in one change, with no negatives needed.

Step 2, page feed or blog exclusion. The blog is less simple. Looking at the $700, around $100 came from two cost guides ("how much does a new water heater cost") that did produce some quote requests. The other $600 went to repair-yourself and explainer posts. The company moves to a page feed with service pages, area pages and those two cost guides. The $600 is gone. The $100 stays.

Step 3, negatives. That leaves $2,800 on service and area pages, plus $100 on the two guides. Inside that, the search terms show intent problems that page targeting can't touch: "water heater installer training", "water heater installation code requirements", "tankless water heater reviews". Phrase negatives for "training", "code requirements" and "reviews", plus a handful of exact negatives, cut another $250 or so.

Result: the campaign drops from $4,000 to roughly $2,650 of spend on the same pages that were producing leads. Without step 1 and step 2, the company would have needed dozens of negatives to get close, and it would still be adding new ones every week as the blog grows.

What the example doesn't show: saving $1,350 is not the same as earning $1,350. If the campaign is on Smart Bidding, some of that budget will move to other searches, and that's usually what you want. Cleaner data also changes what the bidding learns from, as covered in how junk queries skew Smart Bidding. Watch conversions and cost per lead for a few weeks before you call it a win.

Mistakes that make DSA negatives stop working

Most DSA negative keyword problems come from one of four habits. Each is easy to fix once you see it.

Blocking on the search wording alone. "Water heater not heating" looks informational. For a repair company, it's the best search they can get. Before blocking, check whether the term, or terms like it, ever converted. Blocking on a hunch is how good traffic quietly disappears. The method in adding negatives from search terms without guessing applies to DSA just as much as it does to keyword campaigns.

Using broad negatives for short words. A broad negative "repair" in a company that installs and repairs will block far more than you meant. In DSA, where queries are already varied, prefer phrase and exact.

Forgetting the site changes. New blog posts, new landing pages and new product pages become eligible automatically under "all webpages" targeting. An exclusion set up last year doesn't cover a folder created last month. If you rely on exclusions instead of a page feed, review them whenever the site structure changes.

Treating every bad lead as a query problem. Sometimes the search was fine and the page was the problem: wrong page, slow page, no clear next step. DSA makes this more likely because Google picks the page. If the search terms look reasonable and conversions still don't come, read bad clicks or bad pages before adding more negatives.

What PPCleaner does for DSA, and what it doesn't

PPCleaner handles the negative keyword layer. It does not touch page feeds or exclusions. DSA search terms are reported alongside other Search campaign terms in campaign_search_term_view, so PPCleaner reads them every night, judges which ones were never going to convert, and adds them as negative keywords.

What that covers well:

• The long tail of one-click and two-click searches nobody has time to review by hand.
• New drift as it appears, including terms that show up after Google's reading of a page broadens.
• Safety on the terms that matter. PPCleaner never blocks a term that converted in the last 90 days, so "water heater not heating" stays live if it has been bringing in jobs.
• Easy reversal. Every block can be undone in one click for 90 days. If you'd rather do it by hand, see how to undo negative keywords in Google Ads.

What it doesn't cover, and won't pretend to:

• It won't build your page feed or add URL exclusions. PPCleaner never changes targeting, budgets, bids, ads or keywords. If your careers page is pulling job seekers, PPCleaner will block "plumber jobs" and its variants as they appear. Excluding /careers/ yourself is still the better fix, because it stops the whole family at once.
• It can't see what Google doesn't report. Searches hidden from the search terms report can't be judged or negated by anyone, including us.
• It can't fix a page. If good searches land on a weak page, blocking them would be the wrong answer, and PPCleaner won't try.

The practical split: you set the page layer once and revisit it when the site changes. The negative layer is the part that never ends, because DSA keeps finding new searches every day. That's the part worth automating.