AI search engines rewrite a question into several searches, retrieve candidate pages from a web index, pick the passages that best answer each part, and cite the sources they drew from. A client gets cited when their page is retrievable, contains a clean passage answering the specific sub-question, and is corroborated by other trusted sources. That three-part test is also the simplest way to explain AI visibility to a client.
The question every agency is getting
“Why aren’t we showing up in ChatGPT?” usually arrives on a monthly call, often after the client searched their own category and saw a competitor recommended. Your team needs a clear answer, a diagnosis, and a plan, not a shrug.
The four-stage pipeline
- Interpret and expand: the prompt is split into sub-queries (query fan-out).
- Retrieve: candidate pages are pulled from a search index.
- Select passages: the chunks that best answer each sub-query are chosen.
- Generate and cite: the answer is written and sources attributed.
A client can lose at any stage. Blocked crawlers lose at stage 2. Answers buried in paragraph nine lose at stage 3. No third-party mentions often lose at stage 4.
How the main platforms differ
| Platform | How it sources | What to check for clients |
|---|---|---|
| ChatGPT | Search index plus OpenAI’s own crawler (OAI-SearchBot) | Crawler not blocked by robots.txt or firewall; presence in third-party roundups for “best X” prompts |
| Perplexity | Real-time web search with prominent citations | Freshness, clear dated content, community mentions (Reddit) |
| Google AI Overviews / AI Mode | Google’s own index and ranking systems | SEO fundamentals plus topical depth and first-hand experience signals |
What makes a page citable
- Answer-first passages: a direct answer in the first two or three sentences of each section.
- Specificity: numbers, named processes, real examples, dates.
- Consistency: the same brand facts across site, LinkedIn, directories, and reviews.
- Corroboration: third-party sites mention the brand in the same context.
- Freshness and accessibility: visible update dates; crawlable, fast pages.
Turning this into client deliverables
| Deliverable | What it includes | Billing model |
|---|---|---|
| AI citation baseline | 20–50 buyer prompts across ChatGPT, Perplexity, Google; who is cited instead | One-time audit fee |
| Crawler and access check | robots.txt, CDN and firewall rules for AI crawlers | Part of the audit |
| Passage rewrites | Answer-first restructuring of priority pages | Retainer or per-page |
| Mention plan | Target listicles, communities, and publications to earn citations | Monthly retainer |
How to show it in client reporting
Add a monthly citation log to your existing SEO report: prompt, platform, client cited (yes/no), competitors cited, and cited URL. Track the share of prompts where the client appears over time. Clients understand “cited in 12 of 40 prompts, up from 4” far faster than ranking tables.
For agencies
- Package it: a one-time AI citation baseline, then monthly tracking and optimization.
- What to tell clients: “AI picks passages and trusted brands, not just rankings. We’ll fix both.”
- What it takes internally: a prompt library per client, a tracking method, and time for manual review across platforms. This is where most agency teams run out of capacity.
FAQ
AI systems choose passages and corroborated brands, not ranking positions. Clearer answers or more third-party mentions can beat higher rankings.
Only when it searches the web. Answers from training data usually have none, so brand mentions across the web matter for both.
Not in organic AI answers. Ads are a separate channel.
Offer AI citation audits under your brand →
Want to see how your brand ranks in ChatGPT right now?
Run a free AI Visibility Audit to see where your business appears across ChatGPT, Gemini and Perplexity—and where competitors are being recommended instead.
Get Your Free AI Visibility Audit