Striking-distance sprint

SEOmatic is the SEO agent that works from your real Google Search Console data and makes the fixes on your site. Nothing goes live until you click approve, and every change can be rolled back.

The cheapest traffic on any site is already ranking. A query sitting at position 8 has proven the page can rank and proven people search it; the only missing piece is a push. This skill finds those, sizes them honestly, and stages the specific fix.

Every read here is FREE and reads the workspace's own Search Console data. Tools are already scoped to the connected site, so never ask for a domain.

Prerequisite

Search Console must be connected. If a tool reports it is not, give the user this link and stop: https://app.seomatic.ai/connect/gsc - it is one browser step, then re-ask. Do not substitute guesses for GSC data.

1. Pull the candidate set

gsc_performance action query_page_matrix, 28 days. This returns query-page PAIRS, which is what matters: the same query can be split across two pages, and a query ranking 9th on the wrong page is a different problem from one ranking 9th on the right page.

Filter to positions 11-20 with at least 10 impressions. These are the bands SEOmatic itself uses (STRIKING_MIN_POS/MAX_POS/MIN_IMPRESSIONS in the product), so the skill and the app agree on what "striking distance" means

• a shortlist here matches what the dashboard would show.

• Positions 11-20 is page 2. The page has proven it can rank for the query; one edit reaches page 1. That is the whole thesis.

• Positions 1-10 are a DIFFERENT problem and need a different tool. A page-1 result earning less than half its expected CTR is a snippet failure, not a ranking one - see step 3.

• Beyond 20 is not a push, it is a rewrite. Say so rather than selling it as a quick win.

Use a HIGH limit (100+). These queries have few clicks by definition, so a default click-ranked sample systematically misses them.

2. Size the upside, honestly

For each pair: impressions x (expected CTR at target position - current CTR).

The expected-CTR figure is where most estimates become fiction. Do NOT use a published industry CTR curve; use the SITE'S OWN. gsc_insights action ctr_outliers gives the empirical curve for this property, which differs enormously by brand strength, SERP shape and vertical.

State the estimate as a range, and say what it assumes. "Roughly 40-60 more clicks a month if it reaches position 3" is honest. "+52 clicks" is not.

3. Separate the two failure modes

They look identical in the data and need opposite fixes.

• Ranks fine, loses the click (positions 1-10, at least 50 impressions, CTR under HALF the site's expected curve). The ranking is already won; the title and meta are losing the click. gsc_insights action ctr_outliers applies exactly these thresholds and surfaces them directly. Cheapest fix in SEO: no new content, no links.
• Nearly ranks (positions 11-20, the set from step 1). CTR is roughly normal for the position. Content or authority is the constraint, not the snippet - rewriting the title here changes nothing.

Before recommending a title rewrite for a low-CTR pair, check serp_competitors action features with those queries. If an AI overview or a featured snippet now sits above the result, CTR collapsed for a reason the title cannot fix, and the honest recommendation is different: win the feature, or accept the ceiling. Reporting "low CTR, rewrite the title" here sends the user at an unwinnable target.

4. Read the actual pages

site_pages action crawl on the top 5 candidate pages. Compare what the page says against what the query asks. You are looking for a specific, nameable gap:

• The query asks a question the page never directly answers.
• The title targets a different phrasing than the one ranking.
• The intent is mismatched (a commercial query landing on a blog post).

If the page already covers the query well and still sits at 11, say so. The constraint is probably authority, not on-page, and backlink_profile action summary will tell you whether the site can realistically outrank what is above it. A push that needs links is not a quick win.

5. Present a shortlist, not a dump

Ten ranked opportunities beat a hundred rows. For each:

The last column is the one that matters. "Optimize the page" is not a recommendation. "The title targets 'pricing' but the ranking query is 'cost per user' - lead the title with the query phrasing" is.

6. Offer to stage the work

Staging needs a paid SEOmatic plan (the key carries the agents:act scope). It costs no credits, runs nothing, and always waits for the user's approval. On a free key the staging actions are locked and say which plan includes them: describe the specific fix instead and pass that on.

task_manage action create, type ctr_fix for snippet work or striking_distance for content pushes, passing the target queries. Then:

• Tell the user the cards are staged and where to approve them.
• Approved auto-executable tasks start running within SECONDS, so poll task_manage action get shortly after and report what actually changed. The evidence block carries before and after.
• Never approve on the user's behalf, and never claim a result you have not read back from the task.

What not to do

• Do not invent search volumes. Every number comes from GSC or a named tool.
• Do not present the full matrix. The value is the judgment about which ten matter.
• Do not promise ranking movement with a date. Say what changed and what you will measure.
• Do not call a seasonal dip an opportunity. Check gsc_performance action trend first; a query that dips every year at this time is a cycle.