“Anonymised case study” carries a credibility cost, so we’ll be upfront: this composite reflects patterns from real AIOX customers in the independent-publisher segment. The specific numbers are from a Pro-plan customer who agreed to share data; the narrative shape is generalisable.
The site at the start
Mid-sized independent publisher. 800 published posts spanning 4 years. WordPress + a popular SEO plugin. Mix of news, longer-form analysis, and product reviews in a B2B vertical. Monthly traffic before AIOX: ~120,000 sessions, with around 8% from non-Google sources (Reddit, newsletters, direct).
The pain point: organic Google traffic had been declining 5-10% YoY for 18 months. The publisher suspected AI-overview cannibalisation but couldn’t quantify it. Referral traffic from Perplexity was creeping up (~600 sessions/month) but felt random — they couldn’t predict which articles would be cited.
Days 1-7 — Onboarding
- Day 1: signed up for the Pro plan. Installed the Suite plugin on the WordPress install. Ran AIOX Audit on the homepage and three representative posts. Baseline AI Visibility Score: 34. Audit scored homepage at 41/100 (poor schema coverage; no license signals).
- Day 2-3: bulk-processed the entire archive (800 posts). Took 4 hours wall-clock; used 1.05M tokens total. Token pack purchased ($39 for 500k) on top of the Pro allowance.
- Day 4: configured Content Licensing — default “indexing-and-citation”, with PerplexityBot explicitly allowed, GPTBot set to “indexing-only” (no training), Bytespider and CCBot blocked.
- Day 5: turned on AI Bot Sentinel. Started seeing traffic breakdown by bot in real time.
- Day 6-7: ran AI Forensics on the top 5 topic clusters. Found significant stale-info issues — pricing for products covered 18 months ago, a competitor positioned as “comparable” when they’d since pivoted.
Days 8-30 — First impact
- Perplexity referrals: ~600/month → ~1,400/month (over the first 3 weeks).
- ChatGPT Search referrals: from “not enough data to measure” to 200-300/month.
- AI Visibility Score: 34 → 51.
- The publisher noticed they were now consistently cited for two evergreen topics that hadn’t surfaced in LLM responses before.
Days 30-90 — Targeted content work
With the infrastructure in place, the publisher started using AI Forensics outputs to drive content priorities. Specifically:
- Each week, run AI Forensics on the 3 topic clusters with the lowest Visibility Score components.
- Publish 1-2 corrective / authoritative pieces per week targeting the surfaced gaps.
- Process them immediately, so they’re in the AIOX Capsule layer fast.
Day 90 outcomes
- AI Visibility Score: 34 → 71 (+37 points).
- Combined AI-source referral traffic (Perplexity + ChatGPT Search + others): 600/month → 4,200/month.
- Organic Google traffic: stabilised (no longer declining). Possibly correlation, possibly causation; AI optimisation work likely improved overall site signal too.
- Direct mentions in LLM responses for category-defining queries: from “rarely” to “almost always in top 3 sources.”
- Time invested by the publisher: ~6 hours/week ongoing. Mostly content production driven by Forensics priorities.
What didn’t go as expected
- One LLM (we won’t name) initially showed worse Visibility Score on a few topics — turned out it was caching an old response. Cleared on its next training cycle, 6 weeks in.
- The publisher initially blocked too aggressively. Bytespider blocks were uncontroversial; Google-Extended blocks turned out to be premature since Gemini retrievals overlap. They reverted Google-Extended to “indexing-only”.
- The publisher tried to track conversions from AI-source traffic the same way they tracked Google traffic. Conversion rates were different (AI-source visitors arrive more committed; conversion was higher per visitor). New attribution model needed.
What they’d do differently
- Start with AI Forensics on day 1 instead of day 6. The “what do LLMs already say about us?” data shaped the whole strategy and they wish they’d had it sooner.
- Process the most-trafficked 20% of posts first, not chronologically through the archive. The Pareto principle applies: a small minority of posts get most LLM citations.
- Don’t block Google-Extended unless you’re absolutely certain. Gemini retrieval pipeline overlap is non-obvious.
