Skip to main content

Optimization Playbook: What to Fix First

Five categories, in priority order, each with a real before/after example to demonstrate rather than describe.

Once the live exercise surfaces a pile of gaps, the room needs a consistent way to sort them. These five categories — in this order — are that framework. Notice the order isn't arbitrary: it moves from "can a human understand this" to "can a machine find this" to "does anyone believe this" to "are we actually checking."

1. Content Clarity​

What this covers: Answer-first paragraphs, question-format headings, fuller topic coverage.

The measurement: URL Analysis's Answer-First Structure Score — whether H2s are phrased as the questions buyers actually ask, each followed by a direct 40–60 word answer.

Real before/after to show:

Before: Security & privacy (a generic section label, scoring 3/100 on Answer-First Structure) After: How does ai12z protect customer data? followed immediately by "ai12z encrypts all data with TLS 1.3 in transit and AES-256 at rest, enforces role-based access via SSO (SAML 2.0), and operates under HIPAA, GDPR, PCI DSS 4.0, and CCPA..."

The fix costs nothing to implement — it's a rewrite, not new infrastructure — which is exactly why this category sits first in the priority order.

2. Entity & Authority​

What this covers: Specific names, proof, credentials, citations, trust signals — replacing vague marketing language with named, checkable facts.

The measurement: URL Analysis's Entity Density & Node Specificity score, and its E-E-A-T Trust Signals section.

Real before/after to show:

Before: "All data, both in transit and at rest, is secured using industry-standard encryption protocols." (meaningless to an AI system — no algorithm, no key length, nothing to cite) After: "We encrypt data with TLS 1.3 in transit and AES-256 at rest."

Make the point explicit: "industry-standard" is a phrase an LLM cannot quote back to anyone, because it doesn't mean anything specific. "TLS 1.3" and "AES-256" are facts a model can extract, attribute, and repeat.

3. Technical Readiness​

What this covers: robots.txt crawler permissions, sitemap <lastmod> freshness, llms.txt, page speed, Markdown availability.

The measurement: Agent-Readiness Audit.

Real before/after to show: a real audit scored 68/100 ("Mostly Ready") — strong on robots.txt (95), token size (95), and sitemap (95), but zero on llms.txt (missing entirely) and weak on Markdown availability (20, with 0 of 5 key pages resolving a .md URL). The report didn't just flag the llms.txt gap — it generated a complete, ready-to-publish starter file using the site's own real page titles and descriptions. This is the fastest win in the entire playbook: a Critical-priority, Low-effort fix that requires copying generated text to a URL, not writing anything from scratch.

4. Brand Footprint​

What this covers: LinkedIn, Google Business Profile, YouTube, industry directories, cross-platform consistency, duplicate or stale listings.

The measurement: AI Footprint.

Real before/after to show: a real audit scored 48/100 ("Underrepresented in Key Markets") with zero Wikipedia presence, zero confirmed review-platform listings (G2, Capterra, Trustpilot), and a split domain identity (.com vs. .net for docs) that the report flagged as likely reading to AI systems as two separate, unconfirmed entities. The report's Bio & Description Consistency check went a step further than flagging the gap — it wrote a ready-to-use canonical brand description naming the three product surfaces and the core technical mechanic, specifically because one platform (YouTube) had zero descriptive content at all beyond a URL.

5. Measurement Loop​

What this covers: Re-running saved lists, comparing trend over time, consolidating findings into one action plan.

The measurement: GEO Trend Comparison and the Consolidated Action Plan.

Real before/after to show: a real trend comparison between two periods showed GEO Score rising from 76.0 to 78.0 while IDK rate collapsed from 12.6% to 1.8% — clear proof content fixes worked — but competitive intent simultaneously rose from 1.0% to 5.3%, with 13 new competitor mentions appearing. The lesson to draw out for the room: the measurement loop isn't just there to confirm wins — it's there to catch the next problem before it becomes the dominant one. A team that only re-tests to confirm success will miss exactly this kind of emerging competitive pressure.

Why This Order, Restated​

Content clarity and entity/authority fixes are nearly free — rewrites, not rebuilds — so they come first. Technical readiness is slightly more effort but often has a single highest-leverage fix (like the generated llms.txt) that closes most of the gap in one action. Brand footprint work spans multiple external platforms and takes longer to see resolved, so it comes fourth. And the measurement loop only matters once there's something to measure a second time — which is why it's last, not first.

Last updated on