Executive summary
ChatGPT cites sources through two routes: its training corpus and live browsing via its search partnership. To be named you need pages that are crawlable by its agents, written in clear, self-contained, quotable passages, and backed by enough third-party corroboration that the model treats you as a safe answer. This guide sequences the work from access to quotability to corroboration.
What this helps you decide
Whether a page is likely to be retrieved, quoted and attributed by ChatGPT.
Business problem
ChatGPT now answers buyer questions directly, and if it never names or links your brand you are invisible at the exact moment of consideration. Traditional rankings no longer guarantee you appear in the answer.
Step-by-step process
-
1
Let the right crawlers in
Confirm your robots.txt and firewall allow GPTBot (training) and OAI-SearchBot (live retrieval). Blocking them removes you from consideration entirely. Verify with server logs, not assumptions.
-
2
Answer the question in the first 60 words
Lead each key page with a direct, self-contained answer a model can lift verbatim. Models quote passages that make sense without surrounding context, so avoid pronouns and unresolved references in the opening.
-
3
Make claims specific and attributable
Replace vague marketing with concrete figures, dates and named methods. Specific, checkable statements are safer for a model to repeat and more likely to be cited with a link back.
-
4
Add entity and structured-data signals
Use consistent brand naming, an About page, sameAs links and Organization plus FAQ schema so the model can resolve who you are and connect the claim to your entity.
-
5
Earn third-party corroboration
ChatGPT prefers claims echoed across independent sources. Pursue mentions, listings and reviews on sites the model already trusts so your assertion is confirmed elsewhere.
-
6
Cover the full question cluster
Map the follow-up questions a buyer asks and answer each on a dedicated, well-linked page. Breadth of coverage raises the odds you are the source for at least one sub-question.
-
7
Monitor and iterate
Run the target prompts weekly, record whether you are cited, quoted or absent, and feed misses back into the content. Treat citation share as the KPI, not rankings.
Worked example
Checklist
- GPTBot and OAI-SearchBot are allowed and confirmed in logs
- Each priority page opens with a self-contained 40-60 word answer
- Key claims carry specific figures, dates or named methods
- Organization, About and FAQ signals resolve your entity
- At least three independent sources corroborate your core claim
Common mistakes
- Blocking AI crawlers in robots.txt while expecting to be cited
- Burying the answer beneath brand storytelling the model cannot quote
- Relying only on your own site with no third-party corroboration
30-minute experiment
KPIs to track
- Share of target prompts where your brand is cited
- Number of distinct pages quoted by ChatGPT
FAQs
Does ChatGPT link to sources?
In browsing and search mode it shows and links cited sources. In pure conversational mode it may name you from training data without a link, which is why both retrievability and corroboration matter.
How long until changes show up?
Live retrieval reflects recrawled pages within days to a few weeks. Training-based mentions lag by model release cycles, so prioritise the browsing path for speed.
Recommended next steps
Where this fits - and what's next
The SearchScore path from a problem you feel to visibility you can measure.