Dynamic Search Ads: Setup, Page Indexing and Use Cases

by Francis Rozange | Apr 4, 2026 | Google Ads

Category: Google Ads | Reading time: 26 minutes | Last updated: April 2026

Dynamic Search Ads (DSA) sit in an awkward place in the 2026 Google Ads stack. Performance Max is the campaign Google promotes on every onboarding screen. Responsive Search Ads (RSA) are the only standard text ad format left after Expanded Text Ads were sunset in June 2022. And yet DSAs are still here, still serving, still listed in the campaign creation wizard under “Search > Dynamic Search Ads,” still documented at support.google.com/google-ads/answer/2471185. Rumors of their deprecation have circulated since the 2022 RSA mandate and the 2023 Performance Max push, but as of this writing the format is fully operational and Google has issued no end-of-life notice.

The reason DSAs persist is simple: nothing else in the Google Ads inventory does what they do. Performance Max ranges across Search, Display, YouTube, Discover, Gmail and Maps with no transparency on placement-level performance. Standard Search campaigns require keywords, which means a human has to anticipate the query. DSAs fill the gap between those two extremes. They use your website as the keyword list. They generate headlines from the page Google selects. And on a large catalog or a deep content site, they find search demand that no keyword team would have written down.

This guide is the long-form version of the conversation that decides whether DSAs belong in your account in 2026. What they actually do versus what RSAs and Performance Max do, the four targeting modes, page feed setup with custom labels, the situations where DSAs print money, the situations where they bleed it, and the search-terms hygiene routine that keeps the format honest. Sources are linked inline: Google Ads Help, Search Engine Land, Search Engine Journal, WordStream, Tinuiti, Optmyzr.

What DSAs are and why the format exists

A Dynamic Search Ad is a Search Network ad whose headline and final URL are generated dynamically by Google at auction time, based on the content of your website. You do not write the headline. You do not pick the landing page. You write a description (and now, in 2026, a set of fallback descriptions), you choose what Google is allowed to crawl, and you let the algorithm match queries to pages and write the headline on the fly.

The format launched in 2011 and was, at the time, a workaround for a structural problem: large e-commerce catalogs and content-heavy publishers had thousands of pages that nobody could turn into keyword lists fast enough. A retailer with 40,000 SKUs cannot manually build keyword groups for every product. A publisher releasing 200 articles a month cannot wait for the PPC team to write ad copy before the long-tail traffic shows up. DSAs solved that by inverting the input: instead of “give me keywords and I will match them to ads,” the format said “give me URLs and I will match queries to them and generate the ad.”

The mechanism is two-step. First, Google’s web crawler indexes the URLs you allow, extracts the title tag, the H1, the body content, and the structured data, and builds an internal representation of what each page is about. Second, when a user runs a query, Google matches the query against that internal representation, picks the most relevant page, generates a headline that reflects both the query and the page content, and serves the ad. The user sees a headline that looks hand-written. You see, in your search terms report, a query you never bid on and a landing page you never explicitly mapped to it.

The 2024 update that matters: Google now uses the same large-language-model infrastructure that powers Performance Max asset generation to write DSA headlines. The headlines are noticeably better than the 2018-era output (which often pulled awkward fragments from H2 tags), and the relevance score against the query has improved per Search Engine Journal’s 2024 DSA review. The downside is that the LLM occasionally hallucinates product attributes that the page does not actually claim, which is why the search terms hygiene routine described later in this article is non-negotiable.

DSAs vs RSAs vs Performance Max: the three-way comparison

The three formats overlap in coverage but differ sharply in control. Choosing between them is the single biggest architectural decision in a 2026 Search account.

Responsive Search Ads (RSA) are the standard text ad format. You provide up to 15 headlines and 4 descriptions, you specify the target keywords, you write the final URL. Google mixes and matches your assets at auction time and shows the combination predicted to perform best for the query. You control: keywords, headlines, descriptions, landing page. You do not control: which combination of assets shows on a given query.

Dynamic Search Ads (DSA) are the inversion. Google picks the keywords (in practice, the queries) and the landing page from your site. Google generates the headline from the page content. You control: which URLs are eligible, the description text, the bid strategy, the negative keywords. You do not control: the queries you bid on, the headlines, the landing page selection within the eligible set.

Performance Max is the maximalist option. Google picks the placement (Search, Display, YouTube, Discover, Gmail, Maps), the audience, the asset combination, and the bid. You provide a budget, a conversion goal, a creative pool (text, images, video), and audience signals. You do not control: where ads serve, which queries trigger Search placements, which placements Display fills, or how budget splits across channels. Per Tinuiti’s 2024 Performance Max coverage, the search terms report on Performance Max shows only a sample of triggering queries (typically the highest-volume ones), which means dozens to hundreds of long-tail queries are invisible.

The trade-off curve is straightforward: as you move from RSA to DSA to Performance Max, you gain reach and lose transparency. RSA is the format where every dollar is accountable: you know the keyword, the ad copy, the landing page. DSA is the format where most dollars are accountable: you know the description and the URL pool, but not the headline or the precise query trigger until the search terms report ships. Performance Max is the format where almost no dollar is accountable at the placement level.

The decision rule that holds up in 2026: use RSA for the queries you can predict and the products you can name, use DSA for the queries you cannot predict on a site you trust to be content-rich, use Performance Max when you have a Merchant Center feed and a remarketing audience worth feeding it. Run them together, with brand exclusions on Performance Max and DSA, so the cheap branded clicks land in a dedicated branded RSA campaign rather than getting absorbed into the higher-CPC formats.

The four targeting modes

DSAs offer four ways to tell Google which pages to consider eligible. The modes are not mutually exclusive within an account, but each campaign uses one primary approach and you mix them by running multiple campaigns or multiple ad groups.

1. Use Google’s index of my site (categories)

Google crawls your domain, builds an internal taxonomy of categories it detects (for example, “running shoes,” “trail shoes,” “hiking boots” on a footwear site), and exposes those categories as targetable units. You select the categories you want to bid on, and Google routes queries to pages within those categories.

This mode works on sites where the URL structure is deep enough that Google’s crawler can derive meaningful categories. It fails on shallow sites (homepage plus six landing pages) and on sites where the URL structure does not reflect the content (single-page applications with hash routing, URL patterns that hide the product hierarchy). The category list refreshes as Google re-crawls, which means new product lines appear automatically once they are linked from the navigation.

2. Use my page feed

You upload a CSV of URLs to Google Ads Business Data, and DSAs are restricted to those URLs only. The feed can include custom labels (up to four per row) that segment the URLs into ad-group-targetable groups. This is the precision option: Google only considers URLs you explicitly allow, and the custom labels let you bid differently on different segments of the catalog.

Page feed targeting is the recommended approach for any account with more than a few hundred URLs in scope, because it isolates DSAs from URL drift (stale category pages, low-content tag archives, accidentally indexed staging pages) that the “all web pages” mode picks up by default.

3. Target specific URLs, categories, or page titles

Within an ad group, you create dynamic ad targets using rules: “URL contains,” “URL equals,” “page title contains,” “page content contains,” or “category equals.” Multiple rules combine with AND/OR logic. This mode is more flexible than the page feed (you do not need to maintain a CSV) and tighter than “all web pages” (you can exclude entire URL prefixes).

The pattern most agencies converge on: use page feeds for catalog targeting, use URL-based dynamic ad targets for thematic targeting (for example, “URL contains /guide/” to capture all guide articles), use page-title-based targets for ad-group segmentation when the URL structure does not match the desired groupings.

4. All web pages

Every crawlable URL on the domain is eligible. This is the simplest mode and the most dangerous: any URL that Google can reach is fair game, including URLs you forgot existed (old PR pages, sample blog posts, pagination URLs, search result pages on your own site). The mode is appropriate for a small, well-curated site (under 100 pages, all of them landing-page-quality) and inappropriate for almost everything else.

The default in the 2026 wizard is “all web pages,” which is one of the platform defaults to flip before launch. Most underperforming DSA campaigns fail because they were left on this mode and started bidding on internal search pages or thin tag archives.

Page feed setup with custom labels for ad-group-level targeting

The page feed is the under-used lever that turns DSAs from a blunt instrument into a precise one. Done right, it is the difference between a DSA campaign that scales linearly with catalog growth and one that requires manual ad-group rebuilding every quarter.

The feed format is a CSV with two required columns and several optional columns. The required columns are Page URL and Custom label. The optional columns include additional custom labels (up to four total: Custom label, Custom label 1, Custom label 2, Custom label 3), as well as language-specific overrides. Google’s documentation at support.google.com/google-ads/answer/7166527 covers the schema in detail.

The use of custom labels is what makes the feed worth the maintenance overhead. Each label is an arbitrary string you assign to a URL. Once the feed is uploaded, you create ad groups that target URLs by label. The pattern that works: label by margin tier, label by stock level, label by content type, label by promotional status. A footwear retailer might use Margin-High, Margin-Mid, Margin-Low in the first label column and Stock-In, Stock-Low, Stock-Out in the second. Ad groups then bid more aggressively on Margin-High AND Stock-In URLs and exclude Stock-Out URLs entirely.

The feed update cadence matters. Google re-fetches the feed daily if you set it to auto-update from a hosted URL, or on demand if you upload manually. For a catalog with active stock changes, auto-update is the only viable option, because manual uploads always lag the actual inventory. The feed can be hosted on any HTTPS URL Google can reach, including a Google Sheet exported as CSV, which is the lightweight implementation most teams pick before building a proper feed pipeline.

The detail that breaks teams: the page feed is created in Google Ads Business Data, not inside the campaign. Once created, you attach it to the campaign via the campaign settings under “Dynamic ads,” and then you reference the labels at the ad group level via dynamic ad target rules. Three places, three steps, all of them required. Skipping the dynamic ad target rules means the ad group inherits the campaign default (all feed URLs), which defeats the purpose of the labels.

Setup walkthrough: account-level steps

Setting up a DSA campaign in 2026 takes about 45 minutes if the prerequisites are in place (page feed prepared, conversion tracking active, Smart Bidding ready). The walkthrough below assumes Expert Mode and an account that already has GA4 linked and at least 30 conversions in the past 30 days.

Step 1: Create a new Search campaign. In the Google Ads dashboard, click “+ New campaign,” select “Sales,” “Leads,” or “Website traffic” as the goal (the goal affects bid strategy defaults), pick “Search” as the campaign type, and enter the website URL. Untick “Display Network” if it is checked by default; you do not want a Search campaign serving on Display, and DSAs in particular underperform on Display placements per WordStream’s published benchmark data.

Step 2: Configure the campaign as a DSA campaign. Scroll down to “Dynamic Search Ads settings” (in 2026 this section is collapsed by default), expand it, and enter the domain. Select the language. Then choose the targeting source: page feed, Google’s index, or both. For an account with a maintained feed, select “Use URLs from my page feed only” and attach the feed. For an account without a feed, select “Use URLs from Google’s index of my website” and accept the default broader scope, but flip back to a feed within the first 30 days once you have search terms data showing which pages should and should not be eligible.

Step 3: Set the bid strategy. The 2026 default is Maximize Conversions or Maximize Conversion Value, both of which require conversion data. For a brand-new DSA campaign with no prior DSA history, start with Maximize Clicks for the first two weeks to accumulate query data, then switch to Maximize Conversions or Target CPA once the campaign has 15-30 conversions of its own. Smart Bidding on a cold DSA campaign with no historical data behaves erratically for the first 7-14 days regardless, but starting on Max Clicks at least keeps CPCs predictable while the search terms report fills out.

Step 4: Configure ad groups. Create at least two ad groups per DSA campaign: one for the broadest catch-all targeting and one for high-priority URLs or labels. The catch-all group exists to surface long-tail queries and discover gaps in your keyword campaigns. The priority group exists to bid more aggressively on the URLs that drive most of your revenue. Within each ad group, configure the dynamic ad targets: for the catch-all, use “all web pages” or “URL contains” with a broad pattern. For the priority group, use the page feed labels.

Step 5: Write the descriptions. DSAs do not let you write headlines (Google generates those), but you write the descriptions. Provide at least four descriptions, each tightly aligned with your value proposition, and avoid mentioning specific products by name (because the headline will name the product, and repeating the name in the description wastes character count). Strong DSA descriptions read like brand-level positioning: “Free shipping on orders over $50. 60-day returns. Trusted by 200,000 customers.” Weak DSA descriptions read like product-specific copy that fights with the auto-generated headline.

Step 6: Add negative keywords. The first hygiene step is to apply your account-level negative keyword list (jobs, salary, login, careers, free, tutorial, reviews if irrelevant) and any DSA-specific exclusions. The classic DSA-specific exclusions are your competitor brand names (you do not want DSAs serving on competitor queries with weakly relevant pages) and broad informational terms that match your blog but not your product pages (unless your blog is the conversion driver, which it sometimes is on content-led sites).

Step 7: Add dynamic ad targets exclusions. Separately from negative keywords, exclude URLs you do not want DSAs to serve. Common exclusions: /cart/, /checkout/, /account/, /login/, /search/, any URL pattern matching internal search results, any URL pattern matching staging or test environments accidentally indexed. The exclusions live in the campaign’s “Negative dynamic ad targets” section.

Step 8: Confirm location, schedule, and budget. DSAs respect campaign-level location targeting and ad scheduling. Set the location to match your business footprint, set the schedule to match your business hours if you have evidence that off-hours conversion rate drops materially, set the budget at roughly 2x the median DSA CPC in your vertical times the conversion volume target. WordStream’s 2024 benchmark suggests budgeting at least $30-50 per day for a DSA campaign in retail to give Smart Bidding usable signal density.

Step 9: Launch and wait. The first 24-72 hours are crawl and learning. Google fetches the URLs you allowed, indexes them, and starts matching queries. Resist the urge to optimize in the first week. Read the search terms report daily, add negatives aggressively, but do not pause ad groups, do not change bid strategy, do not adjust the page feed. The system stabilizes around day 14, and that is when meaningful optimization begins.

When DSAs work brilliantly: four strong use cases

DSAs are not a default. They are a specialty tool that excels in a small number of scenarios and underperforms outside them. The four scenarios where the format consistently delivers in 2026 are described below.

Use case 1: large e-commerce catalogs

The original use case and still the strongest. A retailer with more than 1,000 SKUs, deep URL structure, well-written product titles and descriptions, and a Merchant Center feed already in place. The DSA campaign captures long-tail queries (specific colors, specific sizes, specific brands plus product type) that the keyword campaigns missed. The page feed restricts targeting to in-stock, high-margin products. The custom labels segment ad groups by margin tier so bids reflect profitability.

The published Tinuiti case studies show DSA campaigns on accounts of this profile capturing 15-25% of total Search-network clicks at a cost-per-acquisition within 10% of the branded Search campaign average, which is the benchmark for healthy DSA performance. The same accounts running DSAs on shallow catalogs (under 200 SKUs) or sites without page feeds typically see DSA cost-per-acquisition 2-3x the keyword campaign average.

Use case 2: content-heavy sites with evergreen articles

Publishers, content marketing programs, and education sites with hundreds of indexed articles benefit from DSAs in the gap between SEO and paid search. SEO captures the queries the articles rank for. Keyword campaigns capture the queries the team explicitly targets. DSAs capture everything in between: the queries an article ranks on page 2 for, the queries that match an article’s topic but use vocabulary the SEO team did not anticipate, the queries that are too low-volume for keyword campaigns to bother with.

The economics work when the content has a downstream conversion path: a B2B blog where articles drive demo requests, an education site where articles drive course signups, a media site where articles drive newsletter subscriptions. Without a downstream conversion, DSA traffic on content is hard to value, and the format becomes an awareness play that competes with cheaper Display inventory.

Use case 3: gap-filling missing keywords on a mature account

The third use case is the discovery layer on top of a mature keyword campaign. The account has been running for years, the keyword lists are deep, the negatives are tight, and the impression share is plateauing. DSAs run as a parallel campaign with low daily budget ($20-50/day) and the explicit goal of surfacing new search terms. The search terms report from the DSA campaign feeds back into the keyword campaign as new exact-match keywords once a query proves it converts.

This is the use case that Search Engine Land’s 2022 DSA optimization article documented, and it has held up through the Performance Max era because Performance Max does not expose query-level data the way DSAs do. The DSA campaign is, in effect, a paid keyword research tool that pays for itself in conversions while it discovers terms.

Use case 4: new market entry

When a business expands into a new geographic market or a new product category, the keyword team has no localized search data to work from. Translating keywords from the home market produces a list that misses local vocabulary and overlooks queries that do not exist in the source language. DSAs solve this by letting the algorithm match local queries to localized pages without the team having to anticipate the vocabulary.

The pattern: launch a DSA campaign tied to the localized site, run it for 30-60 days with aggressive negative keyword maintenance, and harvest the search terms report into a localized keyword campaign once the data is dense enough. This produces a keyword list grounded in actual local query patterns rather than translated guesses.

When DSAs backfire: the failure modes that drain budget

DSAs fail in predictable ways. Recognizing the failure modes early is the difference between a campaign that pays for itself and one that quietly burns budget for months.

Failure mode 1: thin sites with weak content

The format requires content. Pages with sparse text, image-heavy product pages with no descriptions, single-page sites, sites where the navigation is rendered client-side and Googlebot cannot extract it, sites with auto-generated boilerplate descriptions copied across thousands of products. On any of these, the DSA crawler has nothing to work with, the headlines come out generic or wrong, and the relevance score against the query stays low. WordStream’s 2024 DSA coverage at wordstream.com/blog/ws/2023/07/10/dynamic-search-ads calls this out as the single most common reason DSA campaigns underperform: the website was not ready.

The fix is not “tune the DSA campaign.” The fix is to invest in page-level content quality, write real product descriptions, fix the title tags, ensure the H1 is meaningful and unique per page. None of that is DSA work; all of it is SEO work. DSAs benefit from SEO work in the same way Performance Max benefits from a clean Merchant Center feed: the upstream investment is what makes the downstream campaign viable.

Failure mode 2: irrelevant URLs ranking for queries that do not match

The “all web pages” targeting mode picks up URLs you forgot existed. Pagination URLs, tag archives on a WordPress site, internal search result pages, faceted navigation URLs, abandoned blog categories, the careers page, the privacy policy. When Google’s matcher decides one of these URLs is the best fit for a query, the user sees an ad with a headline pulled from a page that does not represent your business. Click-through rates on these mismatches are sometimes high (because the headline accidentally hits the query), conversion rates are zero (because the landing page is irrelevant), and budget burns.

The fix is two-pronged. First, switch from “all web pages” to a page feed or a tightly scoped URL pattern. Second, build a dynamic ad targets exclusion list that explicitly excludes the URL patterns that should never serve: /page/ for pagination, /tag/ for tag archives, /?s= for internal search, /category/uncategorized/ for orphaned blog categories, the careers and privacy URLs by exact match.

Failure mode 3: branded cannibalization

If a DSA campaign has access to your homepage and your “About” page, it will eventually start serving on branded queries (queries that contain your company name). Those queries would have converted anyway from organic search at zero cost, or from a dedicated branded RSA campaign at $0.30 CPC. The DSA serves them at the broader DSA CPC, often $1.50-3.00, and reports the conversions to itself, which makes the campaign look better than it is and starves the branded RSA campaign of impressions.

The fix is the standard brand-exclusion pattern: add your company name and common variants as exact-match negative keywords on the DSA campaign. Run a separate branded RSA campaign with manual CPC and aggressive ad group structure. Repeat the same exclusion pattern on Performance Max if you have it. This is the same hygiene practice Tinuiti’s 2024 paid search series documents for Performance Max, and it applies identically to DSAs.

Failure mode 4: out-of-stock or discontinued URLs

Without a page feed driven by stock data, DSAs will keep serving on URLs that are out of stock or discontinued. The user clicks, lands on a “this product is unavailable” page, and bounces. The ad still cost money. Quality Score on the URL drops because the landing page experience is poor. Future DSA performance on that URL degrades.

The fix is structural: use a page feed, have the feed reflect actual stock (or at minimum, marketplace availability), and update the feed daily. Alternatively, configure the e-commerce platform to return a 410 Gone status on discontinued URLs rather than a 200 OK with a “not available” message, because the 410 status removes the URL from Google’s eligible set entirely.

Interaction with Smart Bidding

DSAs and Smart Bidding are designed to work together, but the integration has subtleties that the platform documentation does not surface clearly.

Smart Bidding strategies (Maximize Conversions, Maximize Conversion Value, Target CPA, Target ROAS) on a DSA campaign require the same conversion volume baseline as on a keyword campaign: 30 conversions in 30 days for stable Maximize Conversions, 50+ for Target CPA, 75+ for Target ROAS. Below those thresholds, the algorithm has insufficient signal density and bids erratically. The 2026 Smart Bidding implementation has improved on the 2020-era version, but the data-volume floor has not changed.

The interaction that catches teams: Smart Bidding on DSAs uses both the query and the page as features. A DSA campaign with a tight page feed produces more predictable Smart Bidding behavior than the same campaign on “all web pages,” because the algorithm has fewer page-query combinations to learn. This is why moving from “all web pages” to a page feed often produces a Smart Bidding stability improvement that looks like the bid strategy was tuned, when in fact the bid strategy is unchanged and the targeting scope was simply tightened.

Per Search Engine Journal’s Smart Bidding guide, the recommended ramp for a DSA campaign is: weeks 1-2 on Maximize Clicks (no conversion data needed), weeks 3-4 on Maximize Conversions once 15+ conversions have accumulated, week 5+ on Target CPA set to the actual observed CPA from weeks 3-4. Setting a Target CPA below the realized CPA at the switch is the canonical mistake: the campaign loses impressions, the algorithm cannot find conversions at the requested cost, and the spend collapses.

The other Smart Bidding interaction worth flagging: Target ROAS on DSAs requires conversion values that reflect the actual transaction value. If the conversion event reports a flat $1, every conversion is equivalent, and Target ROAS becomes Target CPA in disguise. Confirm conversion values are populated correctly before switching to Target ROAS, otherwise the strategy is meaningless.

Exclusion hygiene: negative keywords, URL exclusions, dynamic ad targets exclusions

DSAs require three layers of exclusions, and confusing one for another is the most common DSA mistake in the wild. Each layer addresses a different leakage path.

Layer 1: negative keywords

Negative keywords on a DSA campaign function exactly as they do on a keyword campaign: they block the ad from serving on queries that contain the negative term. Add the standard universal negatives (free, jobs, salary, login, tutorial, careers), the brand exclusions described above, and any vertical-specific negatives that come up in the search terms report. Negative keyword lists from the keyword campaigns can be shared into the DSA campaign via Tools > Shared library > Negative keyword lists, which avoids duplication.

The detail that catches teams: negative keyword match types on DSAs work the same as on keyword campaigns. Negative broad match excludes queries containing all the words in any order. Negative phrase match excludes queries containing the words in the specified order. Negative exact match excludes only the exact query. Most DSA negative lists should use phrase match by default, with exact match for short or ambiguous terms.

Layer 2: dynamic ad targets exclusions (URL exclusions)

This layer is unique to DSAs. It excludes URLs from being eligible to serve, regardless of the query. The exclusions are configured in the campaign under “Negative dynamic ad targets” and use the same rule syntax as the targets: URL contains, URL equals, page title contains, category equals.

The exclusion list every DSA campaign should ship with: URLs containing /cart/, /checkout/, /account/, /login/, /register/, /search/, /404/, /thank-you/, /confirmation/, /unsubscribe/. Add /tag/ and /category/ if your CMS generates those URLs and they should not serve as landing pages. Add the careers, privacy, terms, and contact URLs by exact match. The list takes 10 minutes to build on a typical site and prevents the most common URL leakage failures.

Layer 3: ad group dynamic ad targets

The third layer is positive: at the ad group level, define which URLs are eligible. This is not technically an exclusion, but it functions as one because URLs not matching the ad group’s targets are excluded from that ad group’s serving. Combined with negative dynamic ad targets at the campaign level, the result is a precise serving scope: only the URLs you positively included, minus the URLs you explicitly excluded.

The pattern that holds up: ship every DSA campaign with all three layers configured before launch. Negative keywords for query-level exclusions, negative dynamic ad targets for URL-level exclusions, ad group dynamic ad targets for positive URL scope. Skipping any of the three layers leaves a leakage path that the search terms report will eventually surface, but you would rather discover it in pre-launch review than after a week of wasted spend.

Search terms report and ongoing optimization

The search terms report is the single most important DSA optimization surface. Unlike keyword campaigns, where the search terms report is a hygiene tool, the DSA search terms report is the primary feedback loop on whether the format is working.

The weekly review pattern: pull the search terms report for the last 7 days, sort by cost descending, scan the top 50-100 terms, classify each as “convert,” “promising but unconverted,” “irrelevant,” or “branded.” Add the irrelevant terms as negative keywords. Add the branded terms as negative keywords (they belong in a branded RSA campaign). For the converting terms, decide whether they should be promoted to a keyword campaign as exact-match keywords (which gives you tighter ad copy control) or left in the DSA (which gives the algorithm room to find adjacent queries).

The promotion decision is the strategic one. A converting query in the DSA report represents demand that nobody anticipated. If you promote it to a keyword campaign, you gain ad copy control and probably a small CPC reduction (because exact match on a known-converting term tends to win the auction at lower bids than the broader DSA matching). If you leave it in the DSA, the algorithm continues to discover adjacent queries that match the same intent. The right answer is almost always: promote the high-volume, high-value converting terms to keyword campaigns; leave the long-tail converting terms in the DSA.

The metrics to watch beyond search terms: impression share (low impression share on a DSA campaign usually means budget cap, occasionally low Quality Score on the page set), CTR (DSA CTR varies widely; 2-5% is normal on retail, 4-8% on branded-adjacent terms, anything below 1% suggests headline-page mismatch), conversion rate by ad group (the lever for bid adjustments and pause decisions), and cost-per-acquisition by ad group versus the keyword campaign average.

The optimization cadence per Optmyzr’s published guidance: daily search terms review for the first 30 days, weekly for the next 60 days, biweekly thereafter. The frequency reflects the rate of new query discovery: most of the unique queries surface in the first 30 days, the rate drops sharply afterward, and a mature DSA campaign produces a stable search terms report that needs less attention than a new keyword campaign.

DSA vs Performance Max in 2026: which one belongs in the account

The 2024 conversation around DSAs deprecation came from agencies that had migrated DSA budget into Performance Max and reported equivalent or better performance. The 2025-2026 conversation is more nuanced: Performance Max won the budget reallocation argument for accounts that had Merchant Center feeds and remarketing audiences, and lost it for accounts that did not.

The decision tree that holds up: if you have a Merchant Center product feed actively maintained, a remarketing audience of 1,000+ users, and a creative pool that includes images and ideally video, Performance Max is the better catch-all and the DSA budget probably belongs in Performance Max. If you do not have one of those three (no Merchant Center, no remarketing, no creative pool), Performance Max will struggle with cold-start and DSAs remain the better choice for query discovery.

The hybrid pattern that many agencies converged on by late 2025: run Performance Max as the primary catch-all on accounts that qualify, and run a small DSA campaign in parallel as the search-terms-discovery layer. The DSA campaign has a tight budget ($20-50/day), a tight page feed, aggressive negative keywords, and the explicit purpose of surfacing query-level data that Performance Max does not expose. The search terms from the DSA campaign feed back into both the keyword campaigns and the Performance Max audience signals.

The case where DSAs decisively beat Performance Max: a content-heavy site with no e-commerce. Performance Max requires conversion goals tied to revenue or qualified-lead value to optimize effectively. A B2B blog with newsletter signups as the conversion does not give Performance Max enough value signal to bid sensibly. DSAs on the same site, with newsletter signup as the conversion goal, optimize cleanly because the format is constrained to Search inventory where intent is explicit.

The case where Performance Max decisively beats DSAs: a multi-channel retailer with strong creative assets. Performance Max can serve image ads on Display, video ads on YouTube, and product ads on Search and Shopping inventory simultaneously. DSAs are Search-only. On accounts where the conversion path includes meaningful Display and YouTube influence, Performance Max captures touchpoints DSAs cannot reach.

The 2024 deprecation rumors and the current 2026 status

Three deprecation rumors circulated about DSAs between 2022 and 2024. None of them have materialized as of this writing.

The first rumor, from late 2022, came from the Expanded Text Ads sunset and the Performance Max launch. The reasoning: Google was consolidating ad formats, DSAs were a legacy format from 2011, surely they would be next. They were not. Google issued no deprecation notice, the format continued to be supported, and new features (LLM-driven headline generation in 2024) were added.

The second rumor, from mid-2023, came from Google’s official guidance encouraging Performance Max for advertisers with feeds. The reasoning: if Google is recommending Performance Max for the DSA use cases, the DSA format is being quietly phased out. It was not. The recommendation was a budget-allocation suggestion, not a format end-of-life notice. DSAs remained available in the campaign creation wizard.

The third rumor, from early 2024, came from an internal Google leak about Search ad consolidation. The reasoning: Google was reportedly considering merging DSAs into a unified Search format. The merger never shipped. As of the 2026 platform, DSAs and RSAs are still distinct campaign sub-types with separate documentation pages, separate reporting, and separate support escalation paths.

The reading from Search Engine Land’s 2024 platform updates coverage and Search Engine Journal’s quarterly Google Ads roundups: DSAs are not on the deprecation roadmap. The format has a clear use case Google has not replaced, the LLM upgrade in 2024 indicates active investment, and the integration with Smart Bidding is being maintained alongside the rest of the Search Network. DSAs in 2026 are a stable, supported, niche format. The risk of building a campaign on the format is low; the risk of building an account architecture that depends on the format being the primary catch-all is moderate (because Performance Max is the format Google is investing in most heavily, and the strategic balance may shift in the next 18-24 months).

Common mistakes

Mistake one: launching DSAs on “all web pages” and walking away. The mode picks up internal search pages, tag archives, and accidental URLs. Cost-per-acquisition is 2-3x the keyword campaign average within a month. The fix is page feed targeting from day one, or at minimum a tight URL-contains pattern with aggressive negative dynamic ad targets.

Mistake two: running DSAs without a brand-exclusion negative list. The campaign starts serving on branded queries that would have converted from organic at zero cost. The fix is exact-match negatives on the brand name and common variants, plus a separate branded RSA campaign for the queries that genuinely belong on paid.

Mistake three: writing DSA descriptions that fight the auto-generated headline. The headline says “Trail running shoes for women,” the description says “Buy our new running shoe collection.” The redundancy wastes character count and confuses the user. The fix is brand-level descriptions: shipping policy, return policy, social proof, value proposition, not product-specific copy.

Mistake four: switching to Smart Bidding before the campaign has conversion volume. The 2026 default bid strategy on a new campaign is Maximize Conversions, which on a cold DSA with no historical data behaves erratically. The fix is starting on Maximize Clicks for the first 14 days, accumulating 15-30 conversions, then switching to Maximize Conversions or Target CPA.

Mistake five: ignoring the search terms report for the first month. The search terms in a new DSA campaign are 60-70% noise in the first two weeks. Without daily review and aggressive negative keyword addition, the campaign trains itself on the wrong queries and the noise becomes structural. The fix is daily search terms review with a 5-10 minute time block built into the morning routine.

Mistake six: treating DSAs as set-and-forget because the format is “automated.” The headline generation is automated. The query matching is automated. The budget allocation is not, the negative keyword maintenance is not, the page feed update cadence is not. DSAs require less ongoing attention than keyword campaigns once stable, but more attention than Performance Max in the first 60 days.

Mistake seven: running DSAs on a site Google cannot crawl effectively. Single-page applications with client-side rendering, sites behind aggressive bot protection that blocks Googlebot, sites with thin content. The fix is upstream: server-side rendering, allowlist Googlebot in any bot-protection layer, invest in real product copy. None of this is DSA-specific work, but DSAs cannot work without it.

Conclusion

Dynamic Search Ads in 2026 are a specialty tool with a narrow, defensible use case. They are not the primary catch-all (Performance Max is, on accounts that qualify for it). They are not the precision tool (RSA is, on queries you can predict). They are the discovery layer between the two, the format that finds search demand neither RSA nor Performance Max would have surfaced, the format that converts a content-rich website into a continuous source of paid traffic without the keyword-research overhead.

The accounts that get value from DSAs in 2026 share a profile: more than 500 indexed URLs, real content per page (not boilerplate), a maintained page feed or the discipline to build one, conversion tracking that reflects actual business value, a willingness to do daily search terms review for the first 30 days. The accounts that do not share that profile should run Performance Max as the catch-all and keep DSAs out of the architecture entirely.

The deprecation rumors will continue. They were wrong in 2022, they were wrong in 2023, they were wrong in 2024, and the 2024 LLM upgrade suggests they will be wrong in 2026 as well. Build on the format if it fits the account profile. Skip it if it does not. The decision is architectural, not philosophical, and the architectural answer is rarely ambiguous when the prerequisites are checked honestly.