Why ChatGPT is its own channel
ChatGPT is now the dominant generative search surface. Unlike Google, which returns ten blue links, ChatGPT returns one answer — usually with a shortlist of two to five brand recommendations embedded inside it. Users rarely scroll further; they act on the answer.
That behaviour changes the commercial stakes. In traditional SEO, being ranked tenth still delivers measurable traffic. In ChatGPT, being recommended fourth usually means being excluded entirely — the model has already written its answer before the user sees you.
How ChatGPT decides what to recommend
ChatGPT makes recommendations using three stacked systems. Understanding them is the foundation of ranking well inside it.
1. Pretraining memory
The base model has been trained on huge volumes of public web text. Brands that appear frequently and favourably in that corpus — industry publications, directories, forums, review sites, long-tail blog coverage — build "baked-in" recall. ChatGPT can recommend them even without searching the live web.
2. Live retrieval via browsing
When ChatGPT uses its browsing/search tool, it queries the live web and reads the top results. The model is biased toward pages it can quickly extract: clean HTML, clear headers, structured data, FAQ blocks, quotable claims. Retrieval is your fastest lever for short-term ranking.
3. Source trust
Not every retrieved source counts equally. ChatGPT weights domains it recognises as authoritative — editorial publications, established brands, domains with consistent, accurate content over time. Thin sites, content farms and contradictory signals are ignored or demoted.
What this means for you
You need to win on all three. Pre-training is earned over years through brand authority; retrieval is earned in weeks through technical and content work; source trust compounds over months through consistency. GEO means working every layer simultaneously.
The eleven signals that drive ChatGPT recommendations
1. Structured data, done properly
ChatGPT parses schema markup aggressively. A site with accurate Organization, Product, FAQPage, Article and BreadcrumbList schema is dramatically more extractable than one without. Most sites have partial or broken schema — fixing it is one of the fastest ranking wins you can make.
2. FAQ style content that answers real questions
ChatGPT quotes FAQ content more often than almost any other format. Not because FAQs are special — but because they are structured as a direct question followed by a direct answer, which is exactly what the model needs to generate its own response. Every important page on your site should have a concise FAQ block.
3. Clear, specific claims
Generic marketing copy ("we're passionate about quality") is unquotable. Specific claims ("we ship within two business days from a Melbourne warehouse") are exactly what ChatGPT will pull into an answer. Write like you are feeding an AI a quotable fact, not filling a brand style guide.
4. Consistent brand naming across the web
ChatGPT builds an internal model of your brand from every mention. If your name is written five different ways across your site, your directory listings, social bios and press, the model's confidence drops. Consistency is a ranking signal.
5. Mentions in the publications ChatGPT trusts
Press coverage, industry blogs, podcast appearances, directory listings, review sites — these are the sources ChatGPT tends to cite. A well-placed mention in the right publication can move your brand from invisible to recommended inside a single model update.
6. Accessible, crawlable HTML
Content rendered only by JavaScript, hidden behind scroll triggers, or locked in tabs that don't populate the DOM on load is invisible to most AI crawlers. Your core content must be in the raw HTML.
7. Robots.txt that permits AI crawlers
Many sites unintentionally block GPTBot, ClaudeBot, PerplexityBot, or Google-Extended. If you are not allowing these agents, you are invisible by design. Check your robots.txt — this is a five-minute fix that unlocks everything else.
8. Unique, defensible content
Thin, duplicated or spun content has zero citation value. ChatGPT is trained to ignore it. A smaller volume of substantive, original content outperforms a larger volume of generic content every time.
9. Page speed and mobile rendering
Slow pages get partial retrievals. A page that takes 8 seconds to render may only give ChatGPT the top 1,000 characters before the timeout — everything below the fold is effectively invisible.
10. Topical clustering
ChatGPT is more likely to cite a site that consistently covers a topic well across many pages than a site with one strong page surrounded by unrelated content. Topic depth matters.
11. Fresh updates
Pages updated recently tend to be retrieved more readily than ones that have sat untouched for three years. A gentle, quarterly refresh of your most important content is a straightforward win.
A 30 day action plan
Week 1. Unblock and audit
- Allow GPTBot in robots.txt.
- Run a schema validator on your top 20 pages.
- Run a ChatGPT visibility test with 15–20 queries that matter to your business.
Week 2. Fix structured data
- Complete Organization and Product schema across all templates.
- Add FAQPage schema to every major category or collection page.
- Ensure the FAQs are real questions customers ask, answered concisely.
Week 3. Improve extractability
- Rewrite core page intros to lead with the most quotable factual claim.
- Break long paragraphs into scannable sections with H2/H3 headers.
- Add a "Why we're different" section with three defensible claims.
Week 4. Build citation signals
- Secure at least two placements in industry publications or directories.
- Standardise your brand name and description across every external profile.
- Re-test visibility and compare to the baseline from Week 1.
What not to do
Do not try to "trick" ChatGPT with hidden content, keyword stuffing, exact-match anchor spam or fabricated authority claims. The model is trained to discount exactly these patterns, and the downside if detected is invisibility.
How to measure ChatGPT visibility
Measurement is the most under-developed part of GEO. At minimum you want to track:
- Query coverage: what percentage of the 20–40 queries that matter to your business return you in the answer?
- Share of voice: of all brand mentions across your target queries, what percentage are yours versus each competitor?
- Placement: when cited, are you the primary recommendation, a secondary mention or a passing reference?
- Linked vs unlinked: a linked citation drives clicks; an unlinked mention still builds brand recall.
Our Essentials and Advanced audits run this measurement for you and benchmark against your direct competitors — you see the entire landscape on one page.