IELTS Speaking Simulation GPT V3 update:

1. Updated with the latest question bank for May-August 2024.
2. Improved stability in random question selection.
3. Better experience by automatically asking if you want to improve your answers after the simulation.

Note: The voice feature is still the old version. The real-time features shown at the GPT-4o launch are not available yet.

(Currently, free users can have 16 conversations every 3 hours, which is sufficient for trial purpose. For full practice, consider subscribing to ChatGPT Plus.)

User Guide

Prerequisites

1. IELTS Speaking Simulation is based on ChatGPT’s GPTs and voice functionality. Currently, the voice feature is only available in the official ChatGPT app. Download the ChatGPT app from the App Store and Google Play.

1. Visit the IELTS Speaking Simulation link. If you can’t interact with it, your account may not have free access to GPTs. To use it, subscribe to ChatGPT Plus.

If you have subscribed to ChatGPT Plus, you can also download the macOS version of the ChatGPT app (requires at least M1 chip) to interact directly on your computer. I think it provides a better experience in voice mode.

• ChatGPT macOS version demo
• The method to access the macOS version of ChatGPT early

Finding the IELTS Speaking Simulation

Method 1: Through the Link

With the ChatGPT app installed, use Chrome/Safari browser to open the IELTS Speaking Simulation link. This will directly open the GPT in the ChatGPT app.

To make it easier to access next time, click on the name at the top of the page, then select “Keep in Sidebar.” This will allow you to access it directly from the sidebar in the future.

Finally, to better simulate a fully English conversation environment, it is recommended to set the main language of SPEECH to English in the ChatGPT application settings.

Method 2: Search

If the link doesn’t open the app directly, you can find it through search.

1. Open the ChatGPT app, pull out the sidebar, click “Explore GPTs”, then click the search button at the top right.

1. In the search box, type ”IELTS Speaking Simulator”, and find the GPT with the information shown in the image.

1. Since the temporary page from the search result doesn’t support the voice feature, click the name at the top of the page and select “Keep in Sidebar.” Then, return to the homepage, find the GPT in the sidebar, and click to use it.

1. Finally, to better simulate a fully English conversation environment, it is recommended to set the main language of SPEECH to English in the ChatGPT application settings.

Daily Use

Note: As of May 2024, the real-time voice feature of GPT-4o is not publicly available. The current version still uses the old version functionality released last year.

1. Find the IELTS Speaking Simulation in the Sidebar and select the module you want to simulate. For first-time use, you need to grant permission for external interface calls; choose “Always allow.” After the questions are drawn, click the voice mode button at the bottom right to start the simulation.

(You can also start the conversation directly by telling it which module you want to simulate. However, sometimes it may not call the interface to draw questions, so it is recommended to manually select the module for more stability).

1. During the conversation, if there is a pause, the current input will be interrupted and sent to ChatGPT. If your speech is not fluent enough (especially in Part 2 & Part 3), you can press and hold the center circle while speaking to ensure continuous recording even during pauses. For Part 2, after being informed of the topic, you need to time yourself for 1 minute before starting to speak.

1. After completing the specified module simulation, the system will automatically score your speaking. You can then follow the system prompts to optimize your answers. Due to the length of the content, each part will be output separately.

1. Finally, according to the system prompts, you can summarize the questions and vocabulary used in the simulation. If you have additional needs, you can input them to get further feedback.

Current Issues

1. The score estimation is for reference only (pronunciation is completely inaccurate because the current voice mode converts speech to text. Future real-time voice conversation with GPT-4o may improve this). The main benefit of this GPT is that it offers an authentic practice experience.
2. Part 1 can sometimes be problematic:
3. When providing revised answers, it might miss questions. For example, Part 1 might only give partial answers, or Part 2 might provide insufficient information. You will need to correct it explicitly in such cases.
4. In rare cases, it might display all the drawn questions at once. Fortunately, V3 is much better than before.

If you encounter other issues, feel free to provide feedback or optimize the prompt yourself (I have shard the prompt and question bank for each version. You can download them by clicking on Releases).

Additionally, the voice mode is relatively unstable (affected by OpenAI and your network). Sometimes it might not connect. Try switching to a more stable network or wait a while before trying again. When the network is unstable, not only might you fail to connect, but you might also find that after speaking for a while, ChatGPT informs you that the current usage is high and asks you to try again. 😂😂😂

Additional Thoughts

1. After the GPT-4 update to version 20240409 in April 2024, the initial success rate of question selection with the Code Interpreter significantly improved. However, a week later, it deteriorated, requiring multiple retries to successfully select questions. After trying various methods without success, I updated the random question selection module from ChatGPT’s built-in Code Interpreter to using external interfaces with Actions. This was my first time creating my own interface, indirectly indicating that ChatGPT’s capabilities are still lacking. I look forward to future improvement with GPT-5.
2. Time flies; it has been nearly six months since this GPT was released. During this period, I have simulated multiple times myself, but honestly, I feel my score has only improved by about 0.5 at most. It hasn’t led to significant leaps in improvement. This is partly because I don’t have an immediate need for the exam and haven’t reviewed each practice session thoroughly. If future AI can actively identify unknown issues based on my practice and provide targeted exercises and feedback, that would be impressive.
3. For me, the more important aspect is gaining confidence and having something to talk about on various topics. I hope to use this when I retake the IELTS exam in two years (my goal is modest—scoring a 7 in speaking is sufficient).
4. In terms of experience, current AI still has a long way to go. Otherwise, I wouldn’t need to write such detailed usage guides. 😂😂😂

Thoughts at the Initial Release(Written in December 2023)