Memuat
Memuat
Watch every step above performed live, with English narration.
0 plays
Class book
Page 1 / 5
Cover: Claude from Zero to Expert Class #41: Web Search: When Claude Searches on Its Own
Click a page or use ◀ ▶ to turn · swipe on mobile · to enlarge
Want to go deeper?
Join one of our courses or ask us directly on WhatsApp.
Picked from the software, topic, and level of “Claude from Zero to Expert Class #41: Web Search: When Claude Searches on Its Own”.
Other ways we can help with Claude.
Learn with a mentor
Scheduled AI & Machine Learning courses with live guidance and a certificate.

Midjourney Course: Create Stunning AI Images from Text
★ 4.9

DALL-E Course: Creating Professional AI Images from Text
★ 4.9

Stable Diffusion Course: AI Art from Beginner to Professional
★ 4.9
Have it built for you
No time to learn Claude? Our team can do the project for you.

AI Image Creation with Stable Diffusion — Unlimited Custom Visual Production
★ 4.9

AI Image & Video Creation with Leonardo AI — All-in-One Visuals for Business
★ 4.9

AI Image Creation with Midjourney — Cinematic AI Art for Business
★ 4.9
Get notified about new Claude classes
At most one email a week. Unfollow anytime with one click.
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Hi, and welcome to the research session of Claude from Zero to Expert. From this episode on, Claude doesn't only answer from memory; it also searches the internet by itself. We'll learn when web search is needed, how to turn it on, how to read answers that come with sources, and how to write the right search request.
Early in this class we learned that Claude's knowledge has a cutoff date. It learned from material up to a certain day, and then it stopped. Web search closes that gap. Claude types its own keywords, opens several pages, reads them, and then writes an answer that says where the information came from. Let's compare. Without web search, Claude answers from memory. With web search, the answer is built from pages it reads right at that moment. As a result, it can answer questions about recent things. For example, the drink trend everyone is talking about this month, or next week's small business bazaar schedule. Most importantly, the answer comes with proof: citations and source links you can open yourself. The price is just a little time. Searching and reading several pages takes longer than answering straight away, usually a few seconds up to a few dozen seconds. How do you know a question needs web search? The first sign: it contains words like latest, this year, or this week. Second, the question is about prices, schedules, or stock. Numbers like these get updated often. Third, it's about rules or policies, such as small business tax rules or food labeling rules. Regulations can be revised at any time. Fourth, it's about a specific place or party, such as an event in your city or a particular supplier. For these things, turn on web search.
Now, how to turn it on. First, open a new conversation, in the web app, the desktop app, or on your phone. Second, click the tools menu near the chat box. It's usually a settings icon or a plus sign. Third, turn on the web search switch. The screens in this video are simulations, and menu names may differ slightly because Claude is updated often. You can check availability for each plan on Anthropic's official pages. Fourth, write your question as usual. Once the feature is on, Claude itself decides when it needs to search and when it can answer from memory. Let's look at an example. Laras from Sunset Coffee wants to know which coffee drinks are trending. She mentions her business and her city, and asks for sources. Look at this gray line. Claude decides it needs to search, and then types its own keywords. Next, Claude reads several pages at once, from food news to industry reports. You can see this process on screen, so you know it's really searching. The result is three trends, each marked with a small number in square brackets. Those numbers point to the sources at the bottom. The answer in this video is an example, not real market data. What actually happens behind the scenes? First, Claude understands the question: what to look for, for whom, and in which region. Second, it writes keywords. Sometimes one search is enough; sometimes it searches several times with different keywords. Third, it opens and reads the pages that look relevant, and skips the ones that aren't useful. Fourth, it writes the answer and marks each piece of information with its source number. Because the process works like this, the quality of the answer depends heavily on our question and on the quality of the pages available online.
Answers with sources need to be read in a special way. Laras continues: of those three trends, which fits Sunset Coffee? She also asks for facts and suggestions to be kept separate. Claude splits the answer in two. Facts from sources get citation numbers, while the suggestion is Claude's own opinion that needs testing with customers. Then Laras asks something people often forget: when was source number three published? It turns out it was published last year. Claude offers to search again, limited to the last six months. This simple question about the date saves Laras from stale data. Make three questions a habit for every source. First, relevant: does the source cover the same thing as our question, not just a similar topic? Second, time: when was the source written or updated? Old news can look new if you don't check. Third, match: does Claude's sentence really match what the source says? Click the citation number, open the page, and read the quoted part. We'll go deeper into fact-checking techniques two episodes from now.
Search results depend a lot on how you ask. Take Laras, who wants to join an expo. The vague request just says expo info. The sharp request says food and drink small business expos. Second, region. Without a region, Claude may bring back expos in other cities. Say Bandung and the surrounding area. Third, time range. The next three months makes Claude ignore events that are already over. Fourth, type of source. Ask Claude to prioritize organizer or local government sites, not reposts that may not be accurate. We can turn those four elements into a template. Start with context: who we are and what business we run. Then the task, starting with the words search the web so Claude knows we really want it to search, complete with topic, region, and time range. The limits say which sources to prioritize and what to ignore. Finally, the format. Ask for a table with a source link column and a source updated date column. These two columns make checking much faster. With that template, Claude searches right away with much more precise keywords. The result is a tidy table, complete with how to register and the sources. Claude even mentions one event it skipped because it's already over. The event names here are just examples. Before registering, Laras still opens the organizer's link to confirm the date and requirements. Web search speeds things up, but the decision stays in our hands.
Web search is very helpful, but it has limits. First, online sources can be wrong, and Claude may end up quoting the mistake. Second, paywalled or login-only pages usually can't be read. A locked industry report won't make it into the answer. Third, ads and promotional content can sound like facts. An article written by a seller will of course praise its own product. Fourth, don't put personal data into a search. Don't ask Claude to search for your customers' full names or phone numbers. To wrap up, four habits before using search results. First, make sure the search line appears on screen. That's the sign Claude really searched instead of guessing. Second, open at least one source for important numbers or dates, such as booth rental prices or registration deadlines. Third, ask for a new search when the sources are old. Fourth, keep sourced facts separate from Claude's suggestions. Facts need checking, and suggestions need testing. With these four habits, web search becomes a research tool you can rely on.
We've learned when Claude needs to search the web, how to turn it on, how to read answers with sources, and how to write a sharp search request. In the next episode we go up a level: deep research, where Claude searches dozens of sources to answer a complicated question. See you there!
Try these on your own after watching to lock in what you learned.
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
Notes are saved with the current video time and are only visible to you.
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.

Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
See our clients →