What Quality Score actually measures
Quality Score (1–10) is Google's prediction of how relevant your ad and landing page are to a user searching your keyword. It combines three sub-scores: expected CTR (how likely your ad is to get clicked), ad relevance (how closely your ad matches the keyword), and landing page experience (how useful and relevant your page is to visitors).
Each sub-score is rated below average, average, or above average. A score of 1–4 means at least two of the three components are weak. A score of 7–10 means Google considers your ad genuinely relevant to the searcher.
- Expected CTR: historical and predicted click-through rate for this keyword and match type
- Ad relevance: how closely your ad copy matches the search query and keyword intent
- Landing page experience: page relevance, load speed, mobile usability, and absence of friction
- Quality Score is not used in real-time auction decisions — Ad Rank is. But QS components feed directly into Ad Rank.
Why Quality Score matters for cost
Ad Rank determines your ad position and the price you pay. Ad Rank = bid × Quality Score component × expected impact of extensions. A higher Quality Score means a higher Ad Rank at the same bid — which means a better position at a lower cost per click.
The cost differential is material. Moving from a Quality Score of 5 to 7 on a keyword with a $2.00 CPC typically reduces it to around $1.60–$1.70. Across a campaign spending $5,000/month, that is $300–$400 in saved spend without touching anything else.
The fastest Quality Score improvements
The highest-leverage improvement is usually tightening ad groups. When an ad group contains ten unrelated keywords, the RSA cannot include all of them relevantly. A single-keyword or tight-theme ad group where the main keyword appears in the headline produces much higher expected CTR and ad relevance scores.
For landing page experience, the two most common problems are slow mobile load time and a mismatch between the ad headline and the page headline. If someone clicks an ad for "project management software" and lands on a generic homepage, the landing page experience score drops — and so does Quality Score.
- Add the target keyword in at least two RSA headlines
- Match the landing page H1 to the ad headline keyword
- Ensure the page loads in under 3 seconds on mobile (check with Google PageSpeed)
- Remove navigation links that let visitors leave without converting
- Ask Claude to pull keywords with QS = 5 and diagnose which sub-score is lowest for each
What to do with low-QS keywords
Keywords with QS below 4 and significant spend are costing you in two ways: directly through higher CPC and indirectly by pulling average account quality down. The right action depends on conversion performance: if the keyword converts, improve its ad and landing page alignment. If it does not convert, pause it.
A common mistake is keeping low-QS keywords because they occasionally produce impressions. Impressions from irrelevant traffic have negative expected CTR, which reduces your predicted score further. Pausing the weakest keywords often improves account-wide QS within two weeks.
Using MCP Ads to diagnose QS issues
Connect your Google Ads account and ask Claude: "Pull all keywords with Quality Score below 5 and more than $30 spend in the last 30 days. For each, show which sub-score is below average and the estimated monthly waste from the QS penalty."
Claude can also suggest specific ad copy and landing page improvements based on the keyword theme and current ad content — turning a QS audit into an actionable rewrite list in one conversation.
Doing this with Claude connected to the account
Everything above assumes you are reading Quality Score one column at a time in the Google Ads UI. With Google Ads connected through MCP Ads, the three sub-scores for every keyword are one question away, and the fix can be drafted in the same conversation. Here are three prompts worth pasting.
Prompt 1: "List every keyword with Quality Score of 5 or below and at least 30 days of impressions. Show expected CTR, ad relevance, and landing page experience for each, and group them by which sub-score is the weakest." Claude calls google_ads_quality_score, which returns the historical QS components per keyword, and then sorts the list for you. As an illustrative example, you might get back a table of 40 keywords where 28 fail on ad relevance, 9 on landing page experience, and 3 on expected CTR — which tells you the ad copy, not the site, is the main problem. This is a read-only call, so nothing changes in the account.
Prompt 2: "For the ad groups where ad relevance is below average, show me the current RSA headlines next to the keywords in that ad group, and propose three new headlines per ad group that include the keyword theme." Claude reads the ad groups with google_ads_list_adgroup_keywords and google_ads_list_ads_in_adgroup, then writes the headlines. If you like them, the follow-up "add those headlines to the RSAs" uses google_ads_add_rsa_asset. That is a write, so Claude shows you exactly which ad and which headline it will add and waits for a yes before touching anything.
Prompt 3: "Which landing pages are attached to the keywords with below-average landing page experience, and how do their mobile speed scores compare to the rest of the account?" This combines google_ads_landing_page_performance with google_ads_keyword_performance. What comes back is a list of URLs ranked by how many low-QS keywords point at them, which is usually a much shorter list than the keyword list itself. Fixing two pages can lift twenty keywords.
The safety default applies throughout: reads never ask for confirmation, writes always do, and anything Claude creates from scratch starts paused. You can also run the whole review as a saved workflow using the Google Ads skills instead of typing the prompts each month.
Where the manual way still wins
The API exposes the three sub-scores as bucketed labels (above average, average, below average), not the 1–10 breakdown that Google computes internally, so Claude cannot tell you that expected CTR moved from a 4 to a 5. It also cannot see the auction-time Quality Score that actually sets your CPC, only the historical, keyword-level version everyone sees in the UI.
Landing page experience is where a human still matters most. Claude can rank pages and read their speed data, but it cannot judge whether the page is persuasive, whether the offer matches the promise in the ad, or whether the form is asking for too much. Read the page yourself before rewriting it. And decide for yourself whether a low-QS keyword that converts profitably is worth keeping — the score is an input, not the verdict.
