Audit Compare

Audit Compare turns two audit runs into a narrative — what changed, why it matters, what to try next. Two raw audit results are noisy (“dimension X went 6→7”); a narrative reads the swing in context and tells you whether it’s signal or jitter.

It’s the same pattern monitors use to write their alert emails, the same engine Fasttrak uses for the draft-vs-live comparison, and an explicit user action from the History tab on any landing page workspace.

Three kinds of compare

Audit Compare runs in three flavors. Same engine, different meaning:

Kind	Baseline	Subject	When to use
Temporal	An older audit on the same URL	A newer audit on the same URL	”Did my last copy edit help?” — diff yourself over time
Variant (draft-vs-live)	Live page audit	Preview audit on a draft	”Will this rewrite move the needle?” — verify before you ship
vs. Competitor	An audit on a competitor URL	An audit on your URL (or draft)	“Where do they out-position us?” — competitive benchmark

The kind is set at create time and shows up as a tag in the Orchestrations → Audit Compare list. UI affordances differ slightly per kind — temporal shows up as “Compare to previous” on the History tab, variant as “Compare to live page” inside the Draft preview, vs.-competitor from the Research surface — but the resulting narrative is the same shape.

What you get back

Each compare run produces:

Overall delta — direction (regression / improvement / mixed / noise) and before/after ratings
Per-dimension deltas — each of the 13 audit dimensions with its own before/after and call-out for the biggest movers
Narrative summary — a few paragraphs of plain prose. “Headline appeal moved from 6 to 9 because the new H1 names the buyer’s role explicitly. Evidence quality didn’t move because the BTF still has only generic logos — try replacing one with a named-customer quote.”
has_signal flag — true when at least one dimension moved beyond the noise floor. Drives the “this comparison is worth reading” pill in list views.

The narrative isn’t auto-generated marketing fluff — the prompt grounds it in the actual quoted text from each audit, so the explanation references why a dimension scored how it did, not just that it moved.

Score-only fallback

If the LLM call for the narrative fails (rate limit, outage, malformed response), Audit Compare falls back to a score-only result. You still get every numeric delta — just without the prose. The list row marks score-only runs explicitly so you know the narrative field is empty by failure, not by design.

You can re-run Audit Compare on the same pair of audit runs. The next attempt regenerates the narrative without re-running the underlying audits.

Where it’s triggered

Surface	Kind	Action
Page workspace → History tab	Temporal	Click Compare to previous on any non-first row
Draft workspace → Preview tab	Variant	Click Compare to live page (requires the draft to be linked to a Landing Page)
Research surface	vs. Competitor	Click Compare between an audited competitor and your own page or draft
Fasttrak	Variant	Bundled in Phase 3 of the Fasttrak orchestration — no extra click
Monitor email	Temporal	Auto-fired on every monitor run that crosses the alert threshold

Cost — 1 audit-unit per compare

The compare itself costs 1 audit-unit (the LLM narrative call). The two audit runs being compared each cost 1 audit-unit when they’re created — but if you’re comparing audits that already exist, you’re only paying for the narrative.

Fasttrak and Monitor-triggered compares roll the cost into their parent operation (Fasttrak’s 10 audit-units already include the variant compare; a monitor’s 1 audit-unit per run already covers the temporal compare against the prior baseline).

Plan availability

Audit Compare is on all paid tiers (Solo Founder, Team, Business). Free can’t trigger compares manually — the History tab on Free hides the action — but Free monitor emails still fall back to score-only.