A new analysis of how ChatGPT sources information suggests that the way the model reasons can radically change which brands and websites get visibility. The study, run by Semrush and covered by Search Engine Land, found that ChatGPT in a higher-reasoning, “Thinking” mode cited very different domains than in a minimal, Instant-style mode, with important consequences for marketers, publishers, and product teams. See the full coverage by Danny Goodwin at Search Engine Land. (Search Engine Land).https://searchengineland.com/chatgpt-thinking-mode-brands-sources-citations-481439
Key findings
– Limited overlap between modes. Only 25.6 percent of cited domains overlapped when the same prompts were run in minimal reasoning and high reasoning mode. That means nearly three in four cited sources changed once the model switched from quick answers to a more deliberative search behaviour.
– More citations and more searches. Citation rates rose from 50 percent in minimal reasoning to 68 percent in high reasoning, and the average number of citations per cited answer climbed from 2.6 to 4.5. High reasoning ran 1,130 web searches across the test set versus 245 searches for minimal reasoning.
– Source mix shifted. Reddit fell from a 15 percent share of citations to 7 percent when the model engaged high reasoning. Other user generated content and review sites dropped from 14.3 percent to 6 percent. By contrast, government and academic sources increased from 1.9 percent to 8.8 percent, and official documentation and support pages rose from 12.4 percent to 17.5 percent.
– Comparison prompts amplified fan-out. During comparative questions, high reasoning averaged 24 sub-queries per prompt versus 5.5 for minimal reasoning, and citations peaked at 9.8 per high-reasoning response compared with 5.8 for minimal responses.
– Early citations persisted more often. In four of the 20 buyer-journeys tested, a brand cited at an early problem-definition stage still appeared at the selection stage under high reasoning. Minimal reasoning showed no such full-journey persistence. High reasoning also tended to reuse the same domains within a single answer more frequently: the same domain appeared multiple times in 51 of 100 high-reasoning responses, versus 26 of 100 minimal-reasoning responses.
– Sector differences. The lift in citation rates varied by vertical. Finance saw the biggest jump with an increase of 28 percentage points, health and lifestyle rose 24 points, B2B SaaS gained 16 points, while consumer tech moved only 4 points despite the model running many sub-queries for those prompts.
Why this matters for businesses
The practical takeaway is simple and uncomfortable for many marketers: quick-answer visibility is not the same as considered-answer visibility. If a brand relies on fleeting search placements, social chatter, or review-site buzz, that presence may vanish when a user asks a deeper, comparison-style question and the model switches into a reasoning mode that prioritises documentation, official pages, and authoritative sources.
That shift rewards businesses that invest in clear, accessible documentation, up-to-date support pages, and authoritative white papers. It also penalises those that depend mainly on forums, social aggregation, or thin content for visibility. In other words, companies that have long treated help centres and developer docs as afterthoughts may now be at a structural disadvantage in AI-driven discovery.
Implications for publishers and SEO
Publishers should take notice too. The analysis suggests that a piece that appears in a quick answer may not survive into more deliberative conversations. For newsrooms and specialist sites, the path to sustained visibility is to produce content that can be discovered across the smaller, subject-specific searches the model runs when reasoning. That means robust citation signals, clear metadata, and content which is referenced by official documentation or academic sources when possible.
There is also a structural point to be candid about. The study was conducted by Semrush and reported on Search Engine Land by Danny Goodwin, and Semrush has a commercial interest in search analytics. Readers should treat the results as valuable but not definitive proof of long term production patterns. For the original Semrush write up see Semrushs blog on reasoning and AI visibility. (Semrush).https://www.semrush.com/blog/chatgpt-reasoning-ai-visibility/
What brands should do now
– Audit your documentation. Ensure pricing, integrations, security, and support information is current, machine readable where possible, and signposted from product pages.
– Treat support and docs as SEO assets. Make them linkable, canonical, and easily indexed.
– Prioritise authoritative references. Where possible, strengthen ties to government, industry, or academic sources that the model appears to favour in high-reasoning mode.
– Test journeys not keywords. Run comparison and buying-journey queries against current documentation to see whether your pages surface across the cascade of sub-queries the model runs.
Methodology and caveats
The results come from 100 prompts across 20 buyer journeys in B2B SaaS, finance, consumer tech, and health and lifestyle. Each prompt was executed once in minimal reasoning mode and once in high reasoning mode, with citation rate, cited sources, and fan-out queries tracked. The dataset is useful but limited in scale, and AI models and their default settings evolve rapidly. For full details and the reporting by Search Engine Land, read Danny Goodwins piece. (Search Engine Land).https://searchengineland.com/chatgpt-thinking-mode-brands-sources-citations-481439
Bottom line
This analysis should be a wake-up call for executives who treat SEO as a purely marketing channel. AI-driven discovery increasingly rewards the companies that own the authoritative answers behind their products. Short term social traction may win headlines; durable visibility in the age of large language models will more often come from documentation, official support, and sector authority. For brand teams that want to be visible at both ends of the buyer journey, that is both an operational challenge and a clear investment priority.
