Now Tracking Claude Sonnet 5.5
Centium now tracks Claude Sonnet 5.5, and every model we track can now search the web. What changed, why now, and what to expect.
every model now searches the web.
Centium tracks seven AI models: ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Mode, and Muse. Claude now runs on Anthropic's latest model, Claude Sonnet 5.5, replacing Claude Sonnet 5, and it now searches the web.
Until now, Centium used a blended approach: most models searched the web, while Claude and Grok answered from what they learned in training. Search in these models has improved to the point where that blend no longer earns its place. Grok moved to search earlier this month, and Claude joins it now.
Every model we track now has live web search. Claude, Grok, ChatGPT, Gemini, and Muse decide for themselves, prompt by prompt, whether a question needs it. Perplexity and Google AI Mode search on every answer.
a more efficient Claude made it practical.
We have tested Claude with web search for months, and on earlier Claude models the problem was bloat. Search filled the model's context window, its working memory for an answer, with far more input than the answer needed: a single answer could run 40,000 to 100,000 input tokens. At the volume a Centium study runs, that made Claude with search slow, expensive, and prone to timing out, so we kept it on training data until the model caught up.
Sonnet 5.5 is that model. Anthropic says it "typically needs far fewer tokens to do the same work" (announcement), and it manages what it finds online far better. In our runs, an answer that searched used a fraction of the input tokens the older models needed. That efficiency is what lets us bring Claude with search into Centium reliably, on every prompt in every study.
newer knowledge, at Sonnet speed.
Sonnet is the default model on Claude. When someone asks Claude a general question, the answer almost always comes from Sonnet, with Opus held back for the heaviest reasoning. That is why Centium tracks Sonnet rather than Opus: it is the version most people actually get, and the better fit for the volume of prompts in a study.
Two changes matter most for what Centium measures.
More recent knowledge. The reliable knowledge cutoff moved from January 2026 to June 2026, so Claude's read on your market, your competitors, and what is current starts five months later than it did on Sonnet 5. Search covers what happened since. See every Claude model's knowledge cutoff.
Search available on every prompt. Claude can now look up anything newer than its training on any prompt in your study, the same as the other six models. The model decides for itself when a prompt needs a search.
what to expect in your results.
Your prompts, your categories, and your methodology are unchanged. What changes is how Claude answers them, so you may see some shifting in the Claude numbers on the first update after the switch.
A new model's tendencies. Each model has its own way of answering, so the brands Claude names and how it describes them can shift with the upgrade.
Newer trained knowledge. Claude's knowledge now runs through June 2026, so the last several months of brands and developments are part of what it knows.
Search results. When Claude searches, what it finds can move its answers, and the pages it cites now show up in your Sources and query fan-out data.
After that first run, your Claude trend lines are comparable again. The other six models are unaffected.
on your next update.
The upgrade is automatic, with nothing to turn on. Brands processed since the switch already reflect Sonnet 5.5 with search, and everyone else picks it up on their next scheduled update.
You can confirm which model produced your current data in the Methodology section of your dashboard, which lists every model alongside its knowledge cutoff and whether it uses live search.
questions, answered.
Yes. Claude and Grok both run with live web search, so every model Centium tracks can search. Claude decides for itself, prompt by prompt, whether a question needs it.
No. The upgrade is automatic. Your next scheduled update runs on the new model with search, and brands updated since the switch already reflect it.
Possibly, on the first update after the switch. A new model has its own tendencies, its training runs through more recent knowledge, and search can bring in results that move its answers, so you may see some shifting in Claude's numbers. After that first run, your Claude trend lines are comparable again. Your prompts are unchanged.
The Methodology section of your dashboard lists each model, its knowledge cutoff, and whether it uses live search. Claude reads Claude Sonnet 5.5 once your data has run on it.
More guides
Now Tracking Meta Muse
Centium now tracks Meta Muse as a seventh model on every plan. What changes in your dashboard, what to expect from your first seven-model update, and how to hide or turn off any model.
Which Writers AI Cites in Your Category
Two new features are live. Influential Voices names the people behind the pages AI cites, and Content Freshness shows how old that content is.
Expanded Sentiment Tracking in Brand Perception
Brand Perception shows the themes AI models use when they talk about you. Each theme now expands to explain what the models actually said, in their words, with sentiment attached.