The conversion copy sweep

Most blogs were written for engagement and never reviewed for conversion. The result is familiar: posts that meander, bury the lede, offer no clear next step, or stash a call-to-action in the sidebar but nowhere in the body where a convinced reader would act on it. A conversion lens applies a small, consistent set of rules — the promise visible in the first hundred words, one primary CTA per post rather than three competing ones, scannable subheads, a specific in-body CTA at the natural decision point, and a reinforcing close. Applying that across two hundred posts by hand is weeks of work. With an assistant and your archive on disk, it’s an afternoon — and it’s the recipe most likely to move actual revenue rather than just rankings.

What you need

• Specter synced to your blog
• An AI assistant you already use
• One clearly defined primary conversion goal (email signup, trial, template download) with its target URL and CTA language

The recipe

1. Pull and declare the goal. Run a Specter pull and write down a single primary conversion goal — a sweep optimizing toward one clear action beats one hedging between three.
2. Run the analyzer (no edits). It scores each post against the CRO checks and proposes prioritized fixes — read-only.
3. Triage. Mark each post’s fixes accept, reject, or manual. The “manual” pile is for posts that need a real rewrite, not a tweak.
4. Run the fixer, dry-run, push. Read every diff, reject anything that broke the voice, then push. Track CTA click-through over the next 30 days.

The prompt

The analyzer diagnoses, read-only:

Audit each post against the stated conversion goal. For each, evaluate: does the
first 100 words state the promise? Is there a well-placed in-body CTA matching
the goal (present / weak / missing)? Is it at the natural decision point (~60%
scroll)? Does the closing paragraph reinforce it? Are there competing CTAs to
other actions — list each, every one is a leak. Are the subheads scannable? Is
there a clear next step? Output a per-post scorecard with prioritized fixes and
an accept / reject / manual decision line. Read-only — modify nothing.

The fixer applies only accepted recommendations:

Apply only the accepted fixes per post: insert the primary CTA at ~60% scroll in
the post's own style; rewrite the closing paragraph so the CTA is the natural
next step; sharpen the lede's first paragraph; remove a named competing CTA
(drop the link, keep the sentence unless it's now meaningless); sharpen opaque
subheads without touching section bodies. Never add a CTA where the analyzer
didn't recommend one. Log before/after for every change.

Cost and time

Pitfalls

• Voice intrusion. A CTA written in marketing-speak inside a sharp, conversational post is jarring. Confirm the fixer matched style before pushing.
• Leave pillar and brand pages out. The about page, pricing, and manifesto posts need different treatment — add them to an exclude list.
• Don’t optimize at the cost of the read. The body still has to earn the reader who isn’t convinced yet; the CTA serves the one who already is.

Where to go next

It layers cleanly with the brand voice enforcer in a single pass, and you can A/B the CTA language itself using the headline A/B loop pattern. Wire up conversion tracking before you push so the lift is measurable.