GEO
7 Mistakes Preventing You from Appearing in AI Responses (And How to Fix Them)
May 15, 2026· Facundo Riccillo· 8 min read
After measuring hundreds of brands across ChatGPT, Claude, and Perplexity during 2025-2026, 7 mistakes repeat. If your brand does not appear in AI responses — and LATAM data suggests 60-80% of SMBs do not — you are likely making several of these.
The mistakes
- 1. Optimizing only on-page (schema, llms.txt) and waiting. Not enough.
- 2. Self-promoting on Reddit. Gets banned, detected as astroturfing.
- 3. Measuring only branded prompts. Cold prompts (without your brand) are what count.
- 4. Not measuring noise/variance. Without baseline, any "lift" can be noise.
- 5. Buying reviews. Detected and penalized by review sites AND by LLMs.
- 6. Generic PR without vertical focus. 3 mentions in authority sites > 30 mediocre ones.
- 7. Not having a point of view. Content that synthesizes what others already say is not citable.
Pattern behind the 7 mistakes: treating the problem as technical when it is about marketing and authority. Schema and llms.txt are 20% of the game. The other 80% is building a brand whose content is worth citing.
Quick 15-min self-test
- 15 cold queries on ChatGPT/Claude/Perplexity, count appearances. <10% = problem.
- Search your brand on Reddit — zero organic mentions by others = problem.
- Search G2/Capterra/Trustpilot — few/no reviews = problem.
- Top 5 niche PR sites in your vertical — presence in any? No = problem.
- Your last 3 content pieces — each with unique data/quote/case? No = problem.
If you fail 3+ of the 5 tests, the classic playbook (add schema, llms.txt) will not move your needle. You need to attack the deeper problem of brand authority and differentiation.
Appearing in AI responses is not a technical problem — it is a distributed brand problem.
— ArgentivaOps