How do I show up in ChatGPT, Perplexity, and Claude?

Showing up in these engines means one URL per real customer question, the answer in the first 80 words, and sources, author, date, and FAQ schema on every page. Then the same questions get tested in each engine on a schedule and corrected when they get you wrong. That loop — publish, test, correct — is the whole method, and it is what a Build Site Digital package does for you. Here is what the process looks like so you know what you are buying.

What the process looks like

1. Twenty to thirty questions are collected from real sources: your inbox, your phone notes, the first message customers send, and the questions your front desk answers daily. We build this list with you in one working session.
2. All 20–30 questions run through ChatGPT, Claude, Perplexity, and Google AI results. The answers are saved, and mentions, recommendations, competitors, accuracy, and cited sources are recorded.
3. Each result is scored 0, 1, or 2 — not mentioned, cited but not named, or named. This is your baseline.
4. The five to ten questions where a small site can realistically win get chosen: local, constrained, specific. Broad ones owned by national brands get skipped.
5. One page per question is written: question as H1, answer in the first 80 words, who it is for and not for, steps or a comparison, sources, four FAQs, author and date.
6. FAQPage JSON-LD matching the visible text is added, the page is confirmed to render in HTML, and it is linked from a page that already gets crawled.
7. After thirty days, the identical 20–30 prompts run again and scores are compared. Any answer where an engine states something wrong about you gets fixed.

Want this done for you?

Start with the free audit. If it identifies worthwhile gaps, we provide a fixed scope for the priority improvements and explain the monitoring options.

Engine-specific notes

• — Perplexity cites aggressively and updates quickly. It is usually the first place a new page shows up.
• — ChatGPT blends memory and live retrieval. Broad questions lean on memory; specific and local questions lean on retrieval.
• — Claude leans on live web and social signals, so it can shift day to day.
• — Google AI Overviews appear on some queries and not others, and the same query can show an Overview one week and not the next.

What not to bother with

• — Keyword density. It does nothing here.
• — Publishing forty thin posts. Ten good answer pages beat forty filler ones.
• — Paid "AI ranking" tools that promise placement. Placement is not for sale.
• — Badges claiming an engine recommends you.

A note on honesty

Models change without notice, and inclusion is never guaranteed. A page that is quoted this month can be replaced next month by a source the engine now prefers. The defensible position is to be the most accurate, clearest source for a question you can genuinely answer, and to keep checking.

What this costs

The work is the audit, ten careful pages, and two full rounds of prompt snapshots. Our packages and what each includes are on the pricing page.

Sources

Confirm crawler names and access rules in each vendor's own documentation — OpenAI, Perplexity, and Google Search Central — before changing robots.txt.

Common mistakes in the first month

Three things waste the first month. Publishing broad category pages that compete with national brands, when the winnable questions were local and constrained. Rewriting existing posts instead of adding new URLs, which leaves the question unmatched and the old structure intact. And changing the prompt list between runs, which makes the comparison meaningless — if you reword a question, you have started a new measurement. Keep the 20–30 prompts frozen for at least three runs. Write new questions down in a second list and add them on a later cycle so the baseline stays intact.