Reading an old tutorial safely

You find a video that teaches one useful AI trick. You open the app on your phone and follow along. Nothing on your screen matches the video.

Most people stop here and decide they have broken something. You have not. The video is old. This lesson shows you how to use it anyway.

Why every screenshot goes stale

These companies put out a new model every one to three months. Old models are then switched off on a published date.

Here is the proof, checked on 9 August 2026. GPT-4o was retired from the ChatGPT app for every plan on 3 April 2026. The whole GPT-5 family is due to shut down on 11 December 2026. GPT-5 itself arrived on 7 August 2025. That is a working life of about sixteen months.

The screens change too. In August 2025 one company removed the list that let you pick a model. Many users objected. Within days the older model came back for paying customers. A video filmed that week was wrong the week after.

So an old tutorial is not a lie. It is a photograph of a screen that no longer exists. The speed of these changes is the reason, and nobody can film fast enough to keep up.

Find the button that does the same job

Here is the whole skill in one line. Look for the button that does the same job, not the button with the same name.

Before you copy any step, stop and ask one question. What job is this step doing? Once you can say the job in plain words, you can go and find it.

Say the video taps something called Search the web. Your app has no button with those words on it. The job is still clear. It lets the model read live pages instead of answering from memory. In most apps that setting sits near the box where you type. Open each small menu beside that box and read every word slowly. Our lesson on how AI uses search and tools explains what that job actually does.

You can also ask the AI to name the job for you:

Here is one step from an old AI tutorial.
"Open the model list at the top and choose the newest model."
Tell me in one line what job this step is doing.
Then tell me what that job is usually called today.
Do not guess about my app. Say so if you are not sure.

A good answer names the job. It says the step is about choosing which model replies to you. A weak answer tells you which corner of your screen to tap. The AI cannot see your screen, so treat that part as a guess.

Here is a second example. An old guide tells you to type think step by step before every question. The job there was to make the model work through the steps slowly. Many apps now have a setting for that instead. Look for a word like thinking, reasoning or extended in the menus near the typing box. Our lesson on what thinking mode really does tells you when the extra wait is worth it.

Search for the task, not the model name

Model names die on a schedule. Tasks do not. So put the task in the search box, and leave the version number out.

Type the task the way you would say it out loud:

how to make an AI read a PDF and answer questions on it
how to turn on the slower thinking setting in an AI app
how to ask an AI to check its own answer

Then check the age of what you find. On a video, look for the upload date near the title. On a written page, look for a date near the top or the bottom. A page with no date at all is a page you should not follow step by step. The same test throws out most of the prompt tricks that do not work.

Three signs a tutorial is already dead

Some tutorials cannot be rescued at all. Three signs tell you to stop reading and go and find a newer one.

The first sign is a model name you cannot find anywhere in your app. Search the app once for that name. If it is gone, the guide was written for a version that has been switched off.

The second sign is a way in that has closed. You will meet one example of this again and again. Older guides show you how to chat with ChatGPT inside WhatsApp from an Indian number, by saving a number in your contacts. That way in has closed. Saving the number again will not bring it back. Following the steps more carefully will not help either. Use the ChatGPT app itself, or open it in your browser.

Be careful not to read that as all AI being gone from WhatsApp. Meta AI is still there. It is still the easiest way in the country to try AI, as which AI app you should use first explains. One route closed. The app you already have did not.

The third sign is a promise of a fixed number of free messages. The companies do not publish such a number. OpenAI’s help pages say only that free use is limited. The app tells you when you reach the limit, and it shows you when the limit resets. Google says the Gemini limit refreshes every five hours until you reach a weekly limit. Google also says these limits may change. So read the message your app shows for a time, not for a count.

The three moves that still work

Under all the redesigns, three habits have not changed at all.

Give context. Say who you are, what you are preparing for, and how much time you have. Our lesson on writing a good request breaks this down properly.

Give your own material. Paste the chapter, the notes or the letter instead of trusting the model to remember it. That habit is the subject of giving it your notes and photos.

Say what a good answer looks like. Length, language, and the shape you want it in.

Here is one request that uses all three:

I am in class 10 at a government school in Barabanki.
My science exam is in three weeks.
Below is the chapter from my textbook.
Make ten short questions with short answers, in simple English.
A good answer is one I can revise in ten minutes.

[paste your chapter here]

These three moves worked in 2022, when ordinary people first got these tools. They still work in 2026. No redesign has touched them, because they are about your thinking, not about the app.

Keep your own notes, and read the help page first

Screenshots go stale faster than anything else. Your method does not. So keep a short note for each task you do often, and write it about the thinking, not the buttons.

Task: write a leave application
What I give it: my name, the dates, the reason, the school name
What I ask for: one page, simple English, formal tone
What good looks like: I can copy it and only change the date

That note survives every update. Next month the button may move. Your four lines still tell you exactly what to do.

Keep the note where you will find it again. The notes app on your phone is fine. A cheap notebook is better if the phone is shared at home, because nobody else can delete it. One page for each task you do often is plenty.

When you do need to know where something lives, go to the tool’s own help pages first. Go there before you go to a video. The company updates those pages when it changes the tool, because it has to.

To find them, search the tool’s name with the word help. Then check that the web address belongs to the company itself. OpenAI keeps its help pages at help.openai.com, and Anthropic keeps Claude’s at support.claude.com. Google keeps its Gemini help inside its own support site. Our lesson on what to do when the app does not work uses those same pages.

When it still does not match

Sometimes you will do all of this and still not find the button. That is a normal ending, not a failure.

The people who build these tools change them faster than anyone can write them down. Their own screenshots go out of date. Teachers, video makers and course writers are all behind, and so are the companies. The confusion belongs to the industry, not to you.

You are not slow, and you are not late. A person in a city office is looking at the same changed screen this week. The difference is only that they have stopped blaming themselves for it.

Close the video. Open the app, and tell it what you want in plain words. That is nearly always faster than following an old guide. Next you will look at running a model on your own machine. Old guides go stale there too, for a different reason.