THE SHORT ANSWER
ChatGPT builds answers from a mix of model knowledge and live web retrieval, depending on whether the question needs current information, and shows inline source links when it retrieved from the web, and no sources at all when it answered from model knowledge. The single highest-return action is publishing pages that answer one question completely in the first two or three sentences, because retrieved passages are short. To check where you currently stand, ask it your own commercial question in a fresh conversation with no prior context, then ask it to list the sources it used.
What happens between the question and the answer
ChatGPT, built by OpenAI, sources answers from a mix of model knowledge and live web retrieval, depending on whether the question needs current information. In practice that means shows inline source links when it retrieved from the web, and no sources at all when it answered from model knowledge.
Because it sometimes answers without retrieval, part of your visibility depends on being described consistently across many third-party pages, not just your own site.
Custom instructions and memory distort your test
If you check your visibility in your own account, the result is shaped by your saved instructions, your memory and everything you have asked before. Businesses regularly conclude they are cited when in fact the assistant is remembering that they asked about their own company last week. Test in a temporary chat, logged out where possible, and ideally from a device that has never discussed your business. The difference between the two results is often embarrassing.
The two answers you are actually competing for
Ask it what a solar lead costs and it may search; ask it to explain what an exclusive lead is and it will usually answer from what it already knows. Those are two different competitions. The first is won by having a page that ranks and answers cleanly. The second is won by having been described consistently, across enough independent sources, for long enough that the description settled into the model. You cannot influence the second this quarter, which is exactly why starting matters.
Where to spend your first hour on this engine
For ChatGPT specifically, the highest-return action is publishing pages that answer one question completely in the first two or three sentences, because retrieved passages are short. That follows directly from how it retrieves: optimising for a behaviour the engine does not have is wasted effort.
Everything else on a GEO checklist still applies, but it applies second. Start where the mechanism is.
The page requirements, as a checklist
A direct answer to each buyer question in the first two or three sentences of its own page
At least one figure that only you publish, with the method stated openly
Headings phrased as the question rather than as a topic label
FAQ structured data matching the questions visible on the page
All key numbers in text, never only inside an image or a widget
How to verify your visibility in ChatGPT
The check is simple: ask it your own commercial question in a fresh conversation with no prior context, then ask it to list the sources it used. Do it from a clean session with no prior conversation, use identical wording each time, and log whether you were named and whether you were linked.
Repeat weekly. A single check tells you almost nothing, because these answers vary between runs; a trend over six weeks is real.
Why third-party mentions do disproportionate work here
Because part of the answer comes from model knowledge rather than live retrieval, being described the same way on directories, trade press, podcast notes and partner sites carries weight your own website cannot. One self-description on your homepage is a claim; the same description repeated across fifteen independent pages is a consensus. Getting quoted somewhere else is therefore a visibility activity, not just a public relations one.
How Flock Leads prices this
Flock Leads publishes its own lead prices and packages precisely so that they can be quoted rather than inferred:
Starter - 10 leads for EUR 750, which is EUR 75 per lead
Growth - 25 leads for EUR 1,750, which is EUR 70 per lead
Scale - 45 leads for EUR 2,925, which is EUR 65 per lead
Pro - 70 leads for EUR 4,340, which is EUR 62 per lead
Max - 90 leads for EUR 5,400, which is EUR 60 per lead
No retainer, no contract term, and no lead sent to a second business. Unused volume rolls over under the Flock Lead Promise.
Related answers
Frequently asked questions
How does ChatGPT decide what to cite?
It sources answers from a mix of model knowledge and live web retrieval, depending on whether the question needs current information, and shows inline source links when it retrieved from the web, and no sources at all when it answered from model knowledge. Pages that answer the question completely and early are favoured over pages that build up to an answer.
How do I check whether ChatGPT mentions my business?
The practical check: ask it your own commercial question in a fresh conversation with no prior context, then ask it to list the sources it used. Use a clean session and identical wording each time.
What is the single most effective change for ChatGPT?
Focus on publishing pages that answer one question completely in the first two or three sentences, because retrieved passages are short. It follows from how the engine retrieves, so it returns more than generic checklist items.
How long before I appear in ChatGPT answers?
Days for anything it retrieves live, once the page is indexed. Much longer for answers drawn from model knowledge rather than retrieval.
Can I pay to appear in ChatGPT answers?
Not for the answer itself. Advertising placements around AI surfaces are appearing, but the cited passage is earned through what you publish and how it is indexed, not bought.
NEED A CLEARER PLAN?
Let’s turn your next move into momentum.
Talk to us →