A useful answer to a customer's question may already be on a company's website. If a search system does not know that page exists, however, it cannot consider it. A Cloudflare and OpenAI research pilot is testing whether network signals can help search systems discover potentially useful pages.
On 15 October 2026, Cloudflare reported early results from the pilot launched in July. It said the participating cohort covered about 1% of Cloudflare's network and helped OpenAI discover and index hundreds of millions of new pages. These are the operators' reported discovery figures. They are not counts of ChatGPT citations, site visits, or sales for participating websites.
Our broader SEO and AEO guide discusses content strategy. This new result concerns the earlier step of finding URLs. It is not an announced ranking formula for ChatGPT.
A discovered page is several steps away from a customer
Cloudflare says network signals can help identify pages worth fetching, including pages that have actually changed, without repeatedly downloading every unchanged page. Its pilot also tries to filter nonexistent URLs, misleading redirects, and duplicate content before a crawler spends resources on them. See the pilot report.
Discovery is only the start. OpenAI independently evaluates and ranks pages for relevance and usefulness. Cloudflare does not say that every discovered page appears in a ChatGPT answer or receives a referral. It says the next phase will measure whether better discovery produces real business outcomes. Those outcomes have not been reported in this announcement.
The July 2026 pilot announcement set out to test whether freshness and page-change signals improve search discovery. The October report is an early research result, not a generally available guarantee of placement for a website.
Audit one customer question before publishing more pages
A company with many product pages, manuals, or FAQs can start with one question that its sales or service team actually receives. Follow the page that should answer it.
- Can the intended audience reach it? Open the URL without a login. Check that it shows the relevant content, does not redirect every unknown address to the home page, and does not compete with several outdated copies. Review
robots.txtand bot access rules. Do not make internal documents public just to attract an AI crawler. - Does it provide a verifiable answer? Include the product version, service scope, update date, or limitations a customer needs. Ask the person responsible for the work to check the details, and link to original documentation for factual claims. A generic page can be discovered without being the most useful answer.
- What happens after discovery? Track referrals that your analytics can identify, then check whether visitors take a meaningful next step, such as opening a contact page, downloading public information, or submitting a specific enquiry. Separate crawler requests from people making enquiries; they are different measures.
If the site uses Cloudflare, its owner can use AI Crawl Control to inspect AI crawler requests and allow or block individual crawlers. Cloudflare also says a page-discovery signal does not grant access to the page. According to its pilot explanation, the research uses web metadata for search discovery rather than sharing page contents for model training. Owners still decide access policy according to their rights and goals.
The immediate work is to make pages that customers need accurate, intentionally accessible, and measurable. More content published solely in response to this headline will not establish business value. When the pilot reports citation and business outcomes, site owners can then judge whether to change their investment in content.
The illustration was generated with AI for this article. It does not show actual pilot participants.
