How to fix: Ad groups under-built with too few ads or keywords

TL;DR

A Search campaign whose ad groups carry fewer than three Responsive Search Ads and fewer than three semantically tight keywords cannot give Smart Bidding enough signal to optimize, and Quality Score stays artificially low because Google has too few combinations to learn from [1][2]. The fix is structural: rebuild thin ad groups to a minimum of three RSAs (10-15 unique headlines each) plus a themed keyword cluster of at least three matched-intent keywords, then let Ad Strength and Smart Bidding re-stabilize for 7-14 days.

Why it matters

A Responsive Search Ad is a learning surface, not a static creative. Google rotates up to fifteen headlines and four descriptions against each other and lets the auction-time model pick the combination most likely to win the impression and convert the click [1]. With only one or two RSAs in an ad group, Google has nothing to rotate against. The campaign quietly defaults to single-ad delivery, and the ad rotation report shows a flat 100% share for the single ad regardless of segment or query.

The same principle applies to keywords. Google's account-organization guidance explicitly calls for grouping related keywords around a single product, intent, or buyer stage so each ad group has a coherent theme that one ad set can address [2]. An ad group with one keyword leaves Smart Bidding starving for signal — there's no neighboring intent to triangulate from, and Quality Score components (Expected CTR, Ad relevance, Landing page experience) cannot escape a single ad's ceiling. An ad group with twenty unrelated keywords is the opposite failure mode: relevance collapses because one ad set has to speak to many intents at once.

The pattern usually originates from one of three sources. Rushed campaign launches ship a single RSA per ad group as an MVP and never come back to add variants. Legacy single-keyword ad group (SKAG) structures migrated from the Expanded Text Ad era left behind hundreds of tiny ad groups that never had room for three RSAs in the first place. Abandoned campaign maintenance means ad groups inherit pauses and disapprovals over time until only one RSA is left serving, sometimes silently for months. The audit rule catches all three by measuring the percentage of ad groups under the ≥3 ads / ≥3 keywords baseline against the campaign type filter.

Why three of each, not one or ten. Three RSAs give the auction enough combinatorial space to learn — Google's RSA documentation describes the asset library as a combinatorial test pool, and three ads with ten headlines each yields roughly thirty independent unit tests per impression [1]. Three keywords give the ad group enough thematic mass to coexist with a clear creative angle without diluting relevance across intents. Beyond three, the marginal value tapers off until you cross into ad-group-too-broad territory at roughly fifteen to twenty keywords; below three, the structural starvation kicks in and Smart Bidding cannot read the campaign.

Smart Bidding sits on top of this structure. Maximize Conversions, tCPA, and tROAS strategies need enough conversion events to clear the 30-conversions-per-30-days learning gate, and a campaign of thin ad groups generates very few conversion-worthy combinations to feed the model. The rule's high severity reflects this dependency: bidding strategy and creative quality are downstream of ad group construction, so fixing low Quality Score, low Ad Strength, or wobbling tCPA targets without first fixing this is patching symptoms.

How to fix

1. Open the affected ad group (Campaigns → select the campaign → Ad groups → click the ad group name). Confirm enabled-status ad count and enabled-status keyword count at the top of the page. The audit flags ad groups under three of either.
2. Add Responsive Search Ads until at least three are Enabled in the ad group (Ads & extensions → Ads → + → Responsive Search Ad). For each new RSA, supply 10-15 unique headlines and 3-4 descriptions, varying the angle: benefit, social proof, call to action, brand mention, price point. Do not pin headlines aggressively — pinned positions block Smart Bidding from running its own combinatorial test, which defeats the purpose of having three RSAs [1].
3. Audit the keyword set on the Keywords tab. If the ad group covers more than one product, intent, or buyer stage, split it along the natural fault lines so each new ad group carries one tight theme with at least three keywords. Re-attach RSAs to each split with copy that speaks to that theme.
4. Resolve orphan single-keyword ad groups either by merging them into a themed sibling that already has three or more keywords, or by expanding them with two more close variants of the same intent. Pure single-keyword structures from the SKAG era no longer serve Smart Bidding and should be retired.
5. Hold the ad group untouched for 7-14 days after the rebuild so RSA Ad Strength and Smart Bidding can re-stabilize. Do not stack other major changes (budget shifts beyond 20%, bid strategy swaps, target CPA edits) during the window — Google's documentation describes Smart Bidding as re-entering learning under any of those triggers [3]. Re-run the audit at the end of the window to confirm the rule transitions from failed to passed.

Common mistakes

• Spinning up duplicate RSAs to hit the count. Three near-identical ads with the same headlines reshuffled is not three ads to Google's algorithm — RSA Ad Strength scores them as a single combination set and the rotation experiment never starts.
• Pinning every headline. Aggressive pinning recreates the rigidity of the deprecated Expanded Text Ad and prevents Smart Bidding from running the combinatorial test that justifies having three RSAs in the first place. Accept pinning only for legal or brand mandates (brand name in headline 1).
• Inflating the keyword count without theme discipline. Adding twelve loosely related keywords to push past the three-keyword threshold drags Quality Score down by diluting ad relevance. Three tight-themed keywords beat twelve loose ones every time.
• Reading Ad Strength as a performance metric. Google states explicitly that Ad Strength is a diagnostic, not a KPI [1]. Use it to triage which ad groups need rebuilding; do not target Excellent as a goal in itself.
• Fixing this in isolation. Ad group construction interacts with Quality Score, Smart Bidding learning, and budget pacing. After the rebuild, also audit Quality Score components and confirm the bid strategy is matched to current conversion volume — see the related fix articles via the wiki links below.

FAQ

Does this apply to PMax campaigns? No. PMax uses asset groups, not ad groups — see the Performance Max asset completeness rule and article for the equivalent baseline (1 video, 5 images, 5 logos, 5 short headlines, 5 long headlines, 5 descriptions, 5 sitelinks).

What if the ad group is already converting well with one RSA? Adding two more RSAs does not put the converting RSA at risk — Smart Bidding continues to favor the high-performing combination. The added RSAs widen the exploration window so the campaign can find an even better combination over time.

Should I delete the old RSA when I add a new one? No. Build to three live ads. Pause the old one only if its components are rated Below average after 90 days of stable spend; otherwise let it run.

Is three the magic number, or should I push to five? Three is the floor that unlocks meaningful rotation. Five is the practical ceiling — beyond that you get diminishing returns and creative-management overhead grows faster than the rotation benefit.

Does this rule conflict with budget pacing or bid strategy rules? No, it precedes them. Ad group construction is structural; budget pacing and bid strategy are operational. Fix this first, then re-evaluate the others.

Sources

1. Google Ads Help — About responsive search ads. Documents the 3-15 headline and 2-4 description range per RSA, the combinatorial rotation logic, the recommendation to run at least three RSAs per ad group, and the explicit statement that Ad Strength is a diagnostic rather than a KPI — the canonical source for the RSA half of this rule.
2. Google Ads Help — Organize your account with ad groups. Describes the recommended structure of an ad group: grouping related keywords around a single theme so the ads can address that theme directly, with the electronics-store example as the canonical illustration of theme discipline.
3. Google Ads Help — About Smart Bidding. Explains how Smart Bidding strategies (Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value) depend on signal volume and re-enter a learning phase after structural changes — used here for the 7-14 day hold-still window recommendation.
4. Google Ads Help — About Quality Score. The three Quality Score components (Expected CTR, Ad relevance, Landing page experience) referenced in the Why it matters section as the diagnostic signals that thin ad groups suppress.
5. Wikipedia — Google Ads. Background reference on the Google Ads platform, including the ad group and Responsive Search Ad surface area used throughout this article.