AI Visibility Dos & Don'ts (Q3, 2026)
1. Measure Honestly
Do
-
Baseline how AI describes you across ChatGPT, Gemini, Claude and Perplexity. Trace every gap to a specific page or source.
-
Use AI visibility tool that scrape real chat responses. Run prompts repeatedly. One snapshot is noise.
-
Track branded search volume and direct traffic as your offsetting indicators of AI influence.
Don't
- Don't treat AI mention order as a stable ranking. Outputs are probabilistic and personalized.
- Don't promise linear revenue attribution from AI visibility. Zero-click funnels make it nearly impossible.
- Don't ask an LLM how to rank in that LLM. Models have no insider knowledge of their own retrieval.
2. Protect the Foundation
Do
- Keep winning organic rankings. Pages ranking #1 get cited 3.5x more than pages outside the top 20, and sites that lost organic traffic lost AI citations right behind it, averaging -22.5%.
- Rank for the fan-out sub-queries AI generates, not just head terms. Filter GSC for 10+ word queries to see them.
- Keep key pages crawlable, claims in text not images, and robots.txt open to the AI crawlers you care about.
- Prune and consolidate. Merge overlapping articles into answer-dense hubs.
- Stay ruthlessly on topic. Off-topic content dilutes your site's authority on the subject you actually sell.
Don't
- Don't pipe AI content from an engine into your CMS. It's the mechanism, not the quality: of 220+ sites doing this, 54% lost 30%+ of peak traffic, 39% lost 50%+.
- Don't read the early traffic spike as validation. Spam classification runs after indexing. The spike is the phase before the drop.
- Don't scale comparison / alternative / "best" templates or off-topic how-to libraries.
- Don't use fake review schema or refresh dates without real updates.
- Don't treat schema, llms.txt or markdown files as citation silver bullets. Use them as...