Information
Artificial intelligence: the most popular services. — an informational article for technicians, service centers, and users, covering practical details related to choosing, using, and repairing electronics.

For a phone repair shop, the most useful AI services today fall into three groups: general chat assistants (ChatGPT, Gemini, Claude, Microsoft Copilot), an answer engine that cites its sources (Perplexity), and tools that work only on documents you give them (Gemini Notebook, formerly NotebookLM). They help with customer messages, translations, summaries of service documentation and organizing your notes. Diagnosis, manufacturer schematics and measurements remain the technician's job, and customer data has no place in these services without clear rules.

This article first appeared in June 2023, when ChatGPT, Midjourney and MusicGen were the new names everyone was talking about. We updated it in October 2026. The historical part stays in the past tense, and the rest covers what matters for a repair shop now.

What happened between 2023 and today

OpenAI launched ChatGPT on 30 November 2022 and introduced the GPT-4 model on 14 March 2023, which the original version of this article described. Also in March 2023, Midjourney released version V5 of its image generator, which became known for drawing hands much more convincingly. In June 2023, Meta released MusicGen as open source: a model that generates instrumental music from a text description and was trained on licensed music.

Many names have changed since then. Bing Chat became Copilot in November 2023. Google Bard became Gemini in February 2024. In July 2026, Google renamed NotebookLM to Gemini Notebook. The models behind these services change several times a year, so comparing them by version number is not very useful. It is more practical to know what each one is good for and how to protect your data.

Chat assistants: ChatGPT, Gemini, Claude, Copilot

All four work in a similar way: you type a question or upload a file, and the assistant answers in plain language. They can draft, summarize, translate, explain a technical text and read images. The differences have more to do with the ecosystem around each one than with which is "smarter":

  • ChatGPT (OpenAI) is the most widely used. According to Sensor Tower estimates reported by Reuters, the app passed one billion monthly active users in May 2026. It runs on the web, on desktop and on phones.
  • Gemini (Google) connects to your Google account and, if you allow it, to apps such as Gmail and Drive. Some of its answers are grounded in Google Search results.
  • Claude (Anthropic) works with uploaded documents and files, and Anthropic also offers a separate coding tool, Claude Code.
  • Microsoft Copilot is built into Windows, Edge and Microsoft 365. Since August 2026 the same app accepts both personal and work accounts, and the two stay separate.

All of them have a free tier with limits and paid plans. The terms change often, so check them on each service's website before you choose.

Perplexity and Gemini Notebook: answers with sources

When you ask Perplexity a question, it searches the web in real time and answers with numbered citations that link to the pages it used. That is handy when you want to find out quickly whether a phone model has a known issue or what the manufacturer says about a procedure. Citations do not guarantee that the summary is right, so open the source before you rely on it.

Gemini Notebook works on the sources you give it: PDFs, web pages, Google Docs and public YouTube videos. Its answers include citations that point to passages in those sources. For a repair shop, that means you can upload a service manual or the datasheet of a soldering station and ask questions about the text. Upload only documents you have the right to use.

Where AI actually helps in a repair shop

  • Customer messages. A clear note about an estimate, a due date or warranty terms, which you check and send yourself.
  • Translations. Supplier instructions or datasheets in Chinese or another language, with every number checked against the original.
  • Documentation summaries. The main steps from a long manual, with a reference to the page they came from.
  • Internal procedures. A first draft of a check-in or stocktaking checklist that the team then corrects.
  • Learning. Explanations of electronics concepts, which you then confirm in the manual or at the bench. How to combine theory with hands-on practice is covered in our article on the best way to learn gadget repair.
  • Features inside shop software. Some repair shop platforms have started adding call transcription or recognizing products from a photo. How to choose such a program, regardless of its AI features, is explained in our guide to repair shop management software.

Where you cannot rely on AI

Assistants can state a wrong answer with full confidence: an invented supply voltage, a swapped pin, a procedure meant for a different model. In repair work, a mistake like that can kill a board or a battery. Values, schematics and work steps come from the manufacturer's documentation and get confirmed with a measurement. AI can suggest where to look, but the technician makes the diagnosis.

The same goes for generated images. A generator such as Midjourney, which now works both on its website and on Discord, can produce an illustration for a social media post. A generated "before and after" photo, though, is not real work and does not belong in your shop's portfolio.

Customer data: what not to type into a chat

Everything you type or upload to an assistant ends up on the provider's servers. On personal accounts, conversations may be used to improve the models, depending on your settings:

  • In ChatGPT, you can turn off "Improve the model for everyone" under Settings, Data controls. There is also Temporary Chat, which does not appear in your history.
  • For Gemini, Google states that human reviewers read some conversations and asks you not to enter confidential information. Temporary chats and chats held with Keep Activity turned off are kept for 72 hours.
  • For Claude on personal plans, conversations are used for training only if you have allowed it in settings. Incognito chats are not used for training but are kept for 30 days by default.
  • For Microsoft 365 Copilot Chat used with a work account, Microsoft promises enterprise data protection: prompts and responses are not used to train the underlying models. With a personal Copilot account, check the privacy settings in the app.

The practical rule for the shop: do not enter names, phone numbers, IMEIs, passwords, unlock codes or photos of what is on a customer's phone. If you need help with a message, replace the details with "Customer X" and "Model Y". When you genuinely need AI on real data, use a business plan with written terms on storage and training. What happens to the data on a phone while it is being repaired is a separate topic, covered in our article on myths about smartphone repair.

Rules in the European Union

The EU Artificial Intelligence Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024 and applies in stages. From 2 February 2025, its Article 4 required businesses that use AI systems at work to ensure a sufficient level of AI literacy among the staff who work with them. Regulation (EU) 2026/1744, which entered into force in July 2026, rewrote that article: the obligation remains, but businesses now have to take measures that support the development of AI literacy, and no specific level is mandated. For a small repair shop, that can be as simple as a short internal guide: which services we use, for what, which data we never enter and who checks texts before they go out. The GDPR still applies to customer data. If your shop is outside the EU, check your own country's rules, but the common-sense rules about customer data stay the same.

How to start without taking risks

  1. Pick one assistant and use it for two weeks on tasks with no personal data: message templates, translations, manual summaries.
  2. Check the privacy settings and turn off the use of your conversations for training if the service allows it.
  3. Write the shop's rules on one page: which data never goes in, who reviews the texts, where the results are saved.
  4. Compare every technical value you get from AI with the documentation or a measurement before you use it.
  5. After a month, keep only the uses that save time without adding errors.

Bottom line

Back in 2023, artificial intelligence was a novelty for text, images and music. By 2026 it has become an ordinary work tool, where the model name matters less than how you use it. A repair shop gets real value from it: AI cuts down on office work and speeds up learning, as long as you check what it produces and keep customer data out of it.