How ChatGPT Chooses Sources: 9 Proven Ways Australian Ecommerce Businesses Can Get Found
A closer look at how ChatGPT searches the web, reads websites and chooses which sources to show in its answers.
Share
Sasha Kaluar
Digital marketing expert
Australian ecommerce businesses are asking a new question:
How do we get our products or brand mentioned in ChatGPT?
The usual advice is to publish useful content, earn links, collect reviews and get people talking about your brand. That advice is not wrong, but it leaves out an important detail: how does ChatGPT actually find and use the information?
An analysis of ChatGPT’s browser network activity offers some useful clues. It examined the data sent behind the visible answer, including the searches ChatGPT performed, the websites it fetched and the sources it eventually cited.
This was a limited test using one account and mostly commercial and technology-related searches. The exact percentages should not be treated as universal benchmarks. However, the way ChatGPT searched, fetched pages and separated sources revealed several practical lessons for Australian online stores.
LinkedIn posts get pulled into AI Overviews constantly, because LinkedIn is exactly the kind of user-generated platform Google trusts as a source.
The formula that works: a founder states their positioning claim plainly, in the first line of a post something like “Australia’s leading voice on [X].”
That claim only sticks if the LinkedIn profile backs it up: a headline using the same language, a work history that reinforces it, a consistent posting habit, and a real, engaged following. Take any one of those away and the claim looks hollow to readers and to the algorithm.
Relevance + a platform Google trusts + an authoritative profile on that platform = a claim that starts showing up in AI answers.
How ChatGPT chooses sources for different types of questions
ChatGPT can answer some questions using information it already has. Other questions prompt it to search the web for current information.
OpenAI explains that ChatGPT may search the web based on what someone asks, although users can also choose to search manually.
General questions such as “What is ecommerce SEO?” may be answered without visiting any websites. Searches involving current prices, product comparisons, nearby businesses or recent information are more likely to need web results.
This matters because a page cannot be fetched or cited when no web search happens.
Before investing heavily in an article, consider whether the target question needs fresh information. Searches such as “best Australian skincare brands for sensitive skin” or “Shopify SEO consultant Australia” provide a clearer opportunity than a broad definition that ChatGPT may already know.
Rather than guessing how ChatGPT chooses sources, businesses should focus on reading SAEIGHT’s guide to getting shown in ChatGPT and other answer platforms for more examples of commercially useful searches.
One customer question can become many searches
ChatGPT may rewrite a customer’s original question into several more specific searches.
Someone might ask:
Which Australian standing desk is best for a small home office?
Behind the answer, ChatGPT may investigate standing desk prices, Australian delivery options, product dimensions, customer reviews, warranty terms and comparisons between individual brands.
This is often called query fan-out. It means your page does not only need to match the customer’s original wording. It should also answer the related questions that naturally affect the buying decision.
For an ecommerce product or category page, that could include:
- Current price
- Product dimensions
- Materials and specifications
- Delivery areas and costs
- Warranty information
- Returns
- Suitability for different customers
- Comparisons with similar products
- Frequently asked questions
You do not need a separate article for every variation. One well-structured page can answer several related searches without becoming repetitive.
Important information needs to be easy to read
The analysis showed that ChatGPT often looked for exact facts, including prices, product features and plan names.
When that information was clearly displayed on the official website, the website had a better chance of supporting the answer. When pricing or specifications were difficult to access, ChatGPT could turn to a review platform or another third-party website instead.
This creates a strange situation: your business owns the information, but another website becomes the cited source.
Australian ecommerce websites should place essential product and business information in normal, crawlable webpage text. Avoid relying only on images, downloadable PDFs or elements that appear after complicated scripts run.
Google also provides official guidance on JavaScript SEO and explains that JavaScript-related issues can prevent content from being found or displayed correctly.
Your website can still use modern interactive features. Just make sure customers and crawlers can access the information that matters without solving a digital escape room first.
Being fetched, cited and mentioned are different outcomes
A website can be used in three different ways:
Fetched: ChatGPT opens or retrieves the page while researching.
Cited: The page appears as a clickable source supporting a particular statement.
Mentioned: The business or product appears in the answer, even when another website supports the claim.
These outcomes are related, but they are not the same.
Your product page may be cited for an official price or specification. An independent review may be cited for an opinion about whether the product is worth buying. Your brand could also be mentioned without receiving the citation.
Understanding how ChatGPT chooses sources can help Australian ecommerce businesses explain why your own website cannot do all the work.
Your website owns the facts. Other websites strengthen the opinion.
ChatGPT appears to favour official websites for direct information such as pricing, specifications and policies. Recommendations are more likely to draw on reviews, discussions, comparison websites and established publishers.
For ecommerce businesses, this creates two separate jobs.
First, publish accurate and easy-to-read information on your own website. Second, build enough trust beyond your website that customers, publications and relevant communities discuss your products honestly.
Useful third-party signals can include verified customer reviews, supplier listings, media coverage, industry directories, independent product comparisons and genuine community discussions.
Do not manufacture reviews or create promotional forum posts pretending to be customers. That is risky, transparent and unlikely to build lasting trust.
SAEIGHT covers this in more detail in How to Get AI to Recommend Your Business.
One strong page can beat a collection of thin pages
Publishing dozens of similar pages is unlikely to create dozens of opportunities in the same answer. Search systems may group or remove repeated results from one domain.
A stronger approach is to create one useful page for each major customer need.
For example, an Australian furniture store may benefit more from a detailed standing desk category guide than twenty short articles targeting slightly different versions of “best standing desk Australia”.
A strong page should give the customer enough information to understand the options, compare products and decide what to do next. It should also connect to relevant product pages through descriptive internal links.
Local results can be especially competitive
The network analysis also suggested that ChatGPT may return only a small group of businesses for some local searches.
Whether the exact number remains the same or changes, the lesson is clear: local answer results leave less room than a traditional page of blue links.
Australian businesses should keep their location, service areas, contact information, opening hours and customer reviews accurate across their website and business profiles.
For a local example, see SAEIGHT’s guide to AI search optimisation for Monash businesses.
What should Australian ecommerce businesses do?
Start with the pages closest to a sale: product pages, category pages, buying guides, delivery information, returns, warranty details and pricing.
Make sure those pages clearly explain:
- What the product is
- Who it suits
- What it costs
- How it compares
- Where it can be delivered
- What happens after purchase
- Why customers can trust the business
Then review your visibility outside your website. Look at customer reviews, relevant directories, supplier pages, comparison articles and media opportunities.
You can monitor traditional search performance through Google Search Console and review whether ChatGPT, Perplexity or other platforms are sending referral traffic through Google Analytics.
The real lesson
ChatGPT does not simply choose the website with the most articles.
It looks for clear facts it can support, pages it can access and outside sources that help confirm whether a business deserves to be recommended.
For Australian ecommerce businesses, the goal is not to chase a secret ranking formula. It is to make your products easier to find, understand, compare and trust.
That is good for Google, useful for answer platforms and, most importantly, better for the customer trying to decide what to buy.
The research shows how ChatGPT chooses sources depends on the type of question.
Wondering why your competitors appear in ChatGPT and you don’t?
SAEIGHT helps Australian ecommerce businesses improve how they appear across Google, ChatGPT and AI search. We identify where your brand is mentioned, uncover missed opportunities and build a strategy that helps customers find you first.
