Recommendations Engine
How Recommendations Work
After every audit, 99Visibility's rule-based engine analyzes your results and generates specific, actionable recommendations. These aren't generic tips — each recommendation tells you exactly what to change, where to change it, and how long it should take.
How Recommendations Are Generated
The engine compares your audit results against known GEO best practices. It evaluates your mention rate, accuracy issues, content structure, schema markup, freshness signals, and platform-specific gaps to produce targeted fixes. Recommendations are deterministic: the same audit results always produce the same recommendations, so you can trust the consistency.
Recommendation Categories
Content Structure
Fixes related to how your content is organized. AI platforms extract information more effectively from clear headings, numbered lists, FAQ sections, and well-structured pages. Example: "Add an FAQ section to your /pricing page with structured Q&A format."
Schema Markup
Recommendations about the structured data already on your site. Schema is not an AI ranking lever — Google says none is needed for its AI features — so these are parity checks: every fact in your JSON-LD must also appear in your visible text, and must agree with it. Example: "Your Product schema on /pricing says $49; the visible table says $59. Make them match."
Freshness
Flags pages whose facts have gone stale. No engine documents weighting page age, so this is a consistency check, not a ranking lever: an old page that states a superseded price is a contradiction your newer pages have to fight. Example: "Your /features page lists a plan you no longer sell."
E-E-A-T
Concrete credibility checks — a named author, contact details, claims a reader can verify. Google states E-E-A-T is not a ranking factor and no AI engine publishes an authority score, so these are content practice, not a lever. Example: "Add author names and credentials to your blog posts so a reader can see who wrote them."
Platform-Specific
Fixes targeted at a specific AI platform. Different platforms have different behaviors, so some recommendations only apply to one. Example: "Your brand is not appearing on Perplexity for 8 of 15 keywords. Add citation-friendly content with numbered sources and definitive statements."
Accuracy Fix
Recommendations to correct hallucinations and inaccuracies that AI platforms are spreading about your brand. These are typically the highest priority. Example: "ChatGPT reports your pricing as $29/mo (actual: $49/mo). Two pages on your site still state $29 — update them, and confirm /pricing renders prices in raw HTML rather than via JavaScript." See Hallucination Detection for more.
Priority Levels
- High — fix this immediately. High-priority items are usually accuracy fixes (AI is saying something wrong) or major visibility gaps (you're completely absent from a platform).
- Medium — fix this soon. Medium-priority items improve your score but aren't causing active harm. Things like schema that disagrees with your visible text, outdated content, or pages with no named author.
- Low — fix when you have time. Low-priority items are optimizations that provide incremental improvement. Worth doing, but not urgent.
Effort Estimates
Each recommendation includes an effort estimate so you can plan your work:
- Low (under 30 minutes) — quick wins like fixing a robots.txt rule or correcting a date that disagrees with your sitemap
- Medium (30 minutes to 2 hours) — moderate tasks like restructuring a page or adding an FAQ section
- High (2+ hours) — larger projects like creating new content pages or building authority signals
Implementing Recommendations
- Go to the Recommendations page from your sidebar.
- Sort by priority (High first) and filter by effort level if you want quick wins.
- Click a recommendation to see the full details, including which query triggered it and which platform it affects.
- Implement the fix on your website.
- Mark the recommendation as complete in 99Visibility.
- Run a new audit to measure the impact. We recommend batching 3-5 fixes before re-auditing.
For background on why these recommendations work, see How AI Decides Citations.