Why AI citations compound — and why month 1 looks like nothing's happening.
The short answer
AI citations follow an S-curve: near-zero in month 1 (content is still indexing), a trickle in month 2 (roughly 30% of steady-state), and full pace from month 3 onward. Most businesses give up between week 4 and week 6 — right before the inflection. If you ship consistent content for 90 days, the compounding effect is close to inevitable.
Every founding client has the same month-1 panic attack. They're paying $997, they see 30 articles shipping, the mentions-graph is still flat, and they wonder if we sold them a bill of goods. Then day 45 happens.
Here's the shape of what's actually going on beneath the dashboard. Citation compounding is an S-curve, not a line. If you plot "times you're named in ChatGPT" against days since onboarding, you get something that looks like this:
Mention rate across 10 target prompts · typical founding client
Why the curve stays flat for ~30 days
Large language models don't re-index the web on demand. OpenAI refreshes ChatGPT's crawl window roughly every 4–8 weeks. Perplexity is faster but still batched. Google AI Overviews reads your existing SERPs, which themselves take 10–20 days to reflect new content.
So when we ship 30 articles in November, the median article isn't read by an AI model until mid-to-late December. That's not a bug — it's the physics of the channel.
"The worst thing a client can do in week 3 is panic and start directing the content strategy. The signal simply hasn't been crawled yet."
What's happening on the supply side (even when nothing looks different)
- Article velocity. 30 pieces × cited sources × JSON-LD schema = a compounding topical-authority score.
- Backlink placement. 200–400 vetted contextual links land in month 1 alone. Each is a separate crawl trigger for Google.
- Reddit surface area. The agent answers in your target subreddits. ChatGPT weights Reddit at ~5× other UGC sources.
- Entity graph. Your business name starts co-occurring with city + trade + specific services in a way the models can pattern-match.
The inflection point
Somewhere between day 35 and day 55, three things happen within two weeks of each other: a crawl refresh drops, the backlink graph hits critical mass, and the Reddit threads index. Mention rate jumps from ~20% to ~45% in a single dashboard cycle. From there it climbs weekly.
By day 90 the typical client is at 60–75% on their 10 target prompts — well past the three-prompt guarantee threshold.
What this means for your first month
Read the monthly report. Don't micromanage. Don't "try a different angle" because a friend's agency guy said ChatGPT likes pillar pages now. The engine works because it's consistent, and consistency is what creates the S-curve.
If you want to watch something daily that actually moves — watch your Google organic clicks. That's the leading indicator. Citations follow.