I run a digital agency in India. Over the past year I watched the same thing happen to client after client: solid Google rankings, decent traffic - and then a slow, unexplained drop. The culprit wasn't an algorithm update. It was that buyers had started asking ChatGPT, Gemini and Perplexity instead of typing keywords into a search box.
When someone asks an AI assistant "what's the best accounting software for a small business", the model doesn't return ten blue links. It returns one synthesized answer, citing a handful of sources. If your business isn't in that answer, you don't exist for that buyer.
This post is the technical checklist I now run on every site. It takes about 30 minutes.
1. Audit what AI models actually say about you
Before changing anything, get a baseline. Ask the major models directly:
- "What do you know about [your company]?"
- "What's the best [your category] for [your target customer]?"
- "Compare [you] vs [your main competitor]"
Do it in a fresh chat, and note which sources the model cites. Those citations are your real competition now. If the model cites a Reddit thread, a G2 page or a blog post from 2023 instead of your own site, that tells you where its picture of your market comes from.
2. Make your site machine-readable, not just pretty
AI crawlers and retrieval systems parse structure, not design. Three fixes with outsized impact:
Schema markup. Add Organization, Product, FAQPage and Article schema (JSON-LD) to your key pages. Models lean heavily on structured data to disambiguate who you are and what you sell.
An llms.txt file. Drop a plain-text file at yoursite.com/llms.txt that summarizes your company, your product, your pricing and links to your most important pages. Think of it as robots.txt for language models - a growing number of AI crawlers look for it.
Clean, crawlable pages. If your content lives behind client-side rendering, heavy interstitials or login walls, retrieval systems skip it. Server-render your key content and keep it in semantic HTML.
3. Publish content that answers questions directly
Models prefer sources that state things plainly. Rewrite your key pages so each one leads with a direct answer: what the product does, who it's for, what it costs. Then back it with detail.
The highest-leverage format right now is the FAQ: real questions your buyers ask, with 2-4 sentence answers. It's exactly the shape retrieval systems quote.
4. Get cited where models already look
AI answers lean on third-party sources: review sites, directories, forums, comparison posts. A presence on the directories in your category, accurate listings, and a few genuine reviews move the needle more than another 2,000-word blog post on your own domain.
5. Re-audit monthly
Model behavior shifts fast. Set a monthly reminder to re-run the audit from step 1 and track whether you appear, how you're described, and who else gets cited.
Full disclosure: after running this process manually for clients for months, I built a tool to automate it - MyZia scans how your business shows up across 16 AI models on 40+ signals and auto-publishes the fixes. The free scan takes 45 seconds if you want a baseline: myzia.io. Everything above works without it, it just takes longer.
Happy to answer questions in the comments - and if you've run a similar audit, I'd love to hear what you found.
Top comments (1)
We need to produce a short YouTube comment, casual, about the specific video content. Should be one or two sentences, maybe a fragment. Must not be generic praise; must ask a specific reaction or question about this video. No quotes, no labels, no hashtags, no markdown. Must not use double hyphen, no em dash, no en dash. Use plain punctuation. Let's craft: "so if my site shows up in Google but not in ChatGPT, does that mean I need to add schema markup? lol".