GEO Score 0 → 24 in 31 days, in public. What worked, what didn't, what I'd skip.
Indie dev-tool, 8 weeks old. Started at GEO Score 0 — ChatGPT couldn't find me, Perplexity confused me with another tool, Google AI Mode said 'I don't have specific information about this product.'
Day 31 score: 24. Climb broken down:
What worked (in order of impact):
- Rewriting hero copy claim-first (was metaphor-first). +8 points.
- Adding schema.org
SoftwareApplicationwith explicit category. +6 points. - Publishing one 'how to install X for Y use case' blog post that AI engines indexed within 10 days. +5 points.
- llms.txt with 180-word summary. +3 points (smaller than expected — maybe needs more time).
What didn't work:
- A guest post on a low-traffic dev blog. Indexed by Google, ignored by AI engines.
- Adding lots of FAQ schema to existing pages — moved zero needles.
- Reddit comments mentioning the tool. AI engines don't seem to pick them up at meaningful weight (yet?).
What I'd skip next time:
- The FAQ schema explosion. Pruning to 3 high-quality FAQs would have been smarter than scattering 40.
Questions welcome.
4 replies
- Milan Novák
+8 points from a hero copy rewrite is a bigger jump than I expected from that tactic alone. Was the rewrite mainly about adding the explicit category name, or did you change the structure too?
- Leo H.
The Reddit comment point is interesting. My anecdotal read is that Reddit works for ChatGPT (which heavily weights Reddit/HN for dev tools) but not for Perplexity and AI Mode. So it depends which engine you're trying to move. For my category, ChatGPT is the one that matters most for conversions.
- Sara
What does 'claim-first hero copy' look like vs 'metaphor-first'? Can you share before/after even if anonymized? I think I have a metaphor-first hero right now and want to know if it's actually the problem.
The FAQ schema thing is real. I made the same mistake — added 22 FAQ entries thinking more is better. Pruned to 4 very specific Q/A pairs that match exactly what buyers ask, and Perplexity started pulling them verbatim within 3 weeks. The 22-entry version was getting ignored entirely.