Merlin AI Review: Installed for One Feature But Power Users Quickly Noticed the Limits

Merlin AI is often installed casually, usually as a browser extension someone wants to try for summarising an article, understanding a YouTube video or speeding up an email. At first glance, it feels like a practical addition: lightweight, easy to access and available without repeatedly switching between browser tabs.

For many users, that initial impression is positive. Merlin AI does not demand a complicated setup or a long learning process. It integrates into everyday browsing and writing tasks, which is exactly what makes it appealing.

However, as usage deepens, especially for paying users and people who regularly produce long-form content—some limitations become more noticeable. These are not limited to response length. Query deductions, fair-use restrictions, changing model behaviour and privacy considerations also affect whether Merlin works as a primary AI platform or is better kept as a browser-based assistant

Merlin AI: What It Is Meant to Be vs. How It Feels in Practice

Merlin AI is designed as a browser-based AI assistant that works directly inside websites and web applications. Instead of operating only as a standalone chatbot, it places AI assistance over existing workflows such as reading articles, drafting emails, analysing files and preparing social media content.

Users can open the extension using Ctrl + M on Windows or Command + M on Mac. Merlin also places dedicated AI controls inside selected platforms such as Gmail, LinkedIn, YouTube and X. Its current product pages promote access to several model families, webpage and document chat, image generation, research tools, custom bots, Projects and more than 20 visualisation formats.

In theory, this approach is efficient. Users can select text, ask a question about the current page and continue working without manually copying everything into another application. For short and clearly defined tasks, this remains Merlin’s strongest advantage.

The Chrome Web Store currently shows Merlin with a 4.8/5 rating from approximately 8,800 ratings and around 900,000 users. Google also labels the publisher as following recommended extension practices and having a good record without a history of violations. These are meaningful trust signals, although they do not prove that every paid workflow will perform equally well.

In practice, experienced users, particularly those familiar with the official ChatGPT, Claude or Gemini platforms—sometimes report a gap between expectation and output. Complaints include responses that feel shorter, instructions being missed during longer tasks and a workflow that has become more crowded as additional features have been added.

These experiences are not universal. G2 reviewers also praise Merlin for its browser integration, access to multiple language models and ability to summarise webpages and videos without switching applications. Merlin currently holds 4.1/5 on G2, although that score is based on a relatively small sample of 13 reviews

Evaluation: How Merlin AI Performs in Real-World Use

Based on both general usage and specific user complaints, Merlin AI performs unevenly depending on task complexity:

One of the most common complaints from paying users is that Merlin’s responses sometimes feel artificially shortened. Users who upgraded expecting output comparable to an official ChatGPT or Claude subscription have reported receiving less depth or needing several prompts to complete a long task.

This becomes particularly frustrating for tasks such as:

• Writing long posts or articles
• Organising large amounts of structured content
• Expanding detailed explanations
• Analysing lengthy files
• Producing reports with strict formatting requirements

For these use cases, repeated prompting can reduce Merlin’s main advantage. A browser assistant is supposed to remove steps, but that benefit becomes smaller when the user has to repeatedly ask the tool to continue, restore missing instructions or expand incomplete sections.

It is important not to describe this only as a fixed “token limit,” because Merlin does not publish one simple output cap that applies to every model and task. The final response can be affected by the selected model, context size, query type, web access, uploaded material and Merlin’s own integration settings.

The Query Counter Is Not a Simple Message Counter

Merlin advertises up to 102 free queries per day, but one query does not always equal one AI response.

Its published Query Standards assign different deductions to different models. For example, GPT-4o Mini and Claude 3 Haiku are listed at one query per use, while GPT-4o costs 15, Claude 3.5 Sonnet costs 25 and Claude 3 Opus costs 50. Image models can consume 100 or more queries, and enabling live search can double the normal deduction.

The cost of webpage summaries, blog summaries and YouTube summaries can also vary according to the length and language of the source material. Merlin says that its automatic mode chooses a model and deducts queries according to whichever model and mode were selected.

This means the advertised free allowance should not be interpreted as 102 premium-model prompts every day. A casual user choosing lightweight models may complete many small tasks, while a user selecting premium models or web search can use the allowance much faster.

There is also a documentation gap worth noting. Merlin’s Query Standards still list older model versions, while the current login and product pages promote newer choices such as GPT-5, Claude Sonnet 4.5, Grok 4, DeepSeek R1 and Gemini 2.5 Pro. Users should therefore check the live model selector and visible deduction inside their account rather than relying only on the public query table.

Context Drift: When Instructions Quietly Fall Apart

One of the more frustrating issues reported by users involves context instability. In practical terms, this means Merlin may not consistently follow every instruction during a long or complicated task.

For example, a user may ask the system to retain a fixed structure, avoid translating a particular section or preserve a specified writing style. The response may begin correctly and then gradually move away from those requirements.

This pattern creates what can be described as instruction fatigue. Instead of accelerating the workflow, the tool requires the user to repeatedly restate the same rules.

The impact becomes more noticeable in workflows that depend on precision:

1. Content organisation where headings and formatting must remain consistent
2. Multilingual work where translation boundaries are important
3. Academic editing where wording and citations must be preserved
4. Legal or policy documents where missing details can change the meaning
5. Client work where brand tone and approved terminology must remain stable

These problems should not be presented as affecting every user or every model. They need to be tested using a controlled sequence of follow-up prompts.

A useful context-retention test would begin with five fixed instructions and then continue for at least five messages. The reviewer should record exactly which requirement is first ignored and whether selecting a different model changes the result.

Pros and Cons: A More Realistic Breakdown

 Pros

1. Seamless browser integration
   Merlin fits naturally into everyday browsing and reduces the need to copy information between tabs.
2. Very low learning curve
   The extension is easy to install, and the main sidebar can be opened with a keyboard shortcut.
3. Helpful for lightweight tasks
   Short summaries, quick rewrites, basic research and email drafting are its clearest use cases.
4. Access to multiple AI models
   Users can compare responses from different model families without maintaining a separate account for every provider.
5. Large extension user base
   Merlin currently has approximately 900,000 Chrome users and a strong overall Chrome Web Store rating.
6. Useful page-aware assistance
   The extension can work with the webpage, video or document already open instead of requiring the user to paste everything manually.

Cons

1. Output can feel limited during long tasks
   Some paid users report needing multiple prompts to complete detailed content.
2. Weighted query deductions can be confusing
   Premium models, image tools and web search can consume the free allowance quickly.
3. Paid use is not unrestricted in every situation
   Merlin’s Terms include daily and monthly internal cost thresholds.
4. Context retention may weaken during complicated workflows
   Detailed instructions can require repeated supervision.
5. Model behaviour may differ from official platforms
   Selecting a familiar model name does not guarantee an identical interface or output.
6. The growing interface can feel crowded
   Some G2 users say Merlin has become more cumbersome or overwhelming as more functions have been added.
7. Browser activity is processed
   Users should understand the extension’s URL and activity-data collection before using it with sensitive work.

Final Verdict

Merlin AI works best as a convenience-focused browser assistant for users who want quick help with reading, writing and research directly inside a webpage.

Its core value remains easy to understand. A user can open an article, video, email or social media page and ask for assistance without moving the content into another tool. For short tasks, that can genuinely save time.

The overall user picture is mixed rather than entirely negative. Merlin has a 4.8/5 Chrome Web Store score from around 8,800 ratings and a 4.1/5 G2 score from 13 reviews. Trustpilot presents a different view, with a 2.0/5 score from 21 reviews and frequent complaints concerning subscriptions, support and output quality. Trustpilot also notes that the company has not invited reviews there, so its smaller sample may not represent the wider customer base.

For casual users, students, marketers and professionals completing brief browser-based tasks, Merlin can be useful. The free plan is enough to test its summarisation, rewriting and page-aware features before paying.

Power users should be more cautious. Long-form writers, researchers, analysts and anyone processing many large documents should test output length, context retention and query deductions using their actual workflow. They should also understand that Merlin’s paid access remains subject to documented internal usage limits.

Merlin is therefore not simply a poor version of ChatGPT or Claude. It is a different type of product: an aggregation and productivity layer built around browser convenience.

That convenience is valuable when the task is short and the page already contains the necessary context. It becomes less valuable when accuracy, predictable long-form output, confidential data handling or complete control over the original model experience matters more.