Using Claude, the AI Chatbot by Anthropic, as a Line Editor and Critique Group

Using Claude AI as a Line Editor and Critique Group

This is a follow up on The best of times, the worst of times—Artists, Writers and AI, Part 1. The first post outlines how I want to explore the possibilities of AI and AI-assisted tools while maintaining some idea of ethical integrity, while this post focuses on Claude.

I write on an island in an otherworldly time zone without an agent or writing group. Not a perfect set-up for any writer. Neither is writing in an adopted language, or writing outside current fashions and commercial trends. Claude offers feedback on a budget and on-call. It revises my drafts at 3:00am without complaints, and interactions are free of emotional baggage. 

I ask Claude for feedback on my writing almost daily, both for my paid work (technical writing and marketing blurbs) and for my creative writing experiments. I don’t ask Claude to write first drafts or come up with ideas for what to write—I ask it to analyse and interpret what I have already written.

How Claude ‘Works’

You’ll need to ask a much smarter person than me to explain the inner workings of AI. On the surface, AI chatbots process and generate text using language patterns learned from vast amounts of data. They answer your questions by ‘predicting’ the next word—or even just a part of a word—in a sentence, but on a much larger and more complex scale. Each release is based on an ever-growing set of training data*. In essence, AI chatbots are creating content based on probabilities.

Despite that, the results are so close to an intelligent conversation with a smart person that it feels … unreal. Until the conversations turn to hallucinations, a term used to indicate the AI chat bot spews out false information wrapped in perfect sentences.

I have encountered few hallucinations while asking Claude for feedback on my fictional drafts. When is comes to feedback on grammar, punctuation, spelling, and style, Claude might hallucinate from time to time but still finds more mistakes, or other mistakes, than Google Docs or ProWritingAid. It knows when I meant trash chute and not trash shoot, or leaf instead of leave. (Homonyms are a common problem for non-native speakers).

Let’s have a look at two hands-on examples.

Using Claude as a line editor

Line editing happens on an advanced draft—every scene, paragraph, and sentence are scrutinised to improve the style, flow and readability of your manuscript before the focus shifts to copy editing and proof reading. I usually ask Claude for feedback per scene—a scene is collection of events in a complete unit of structural plot with a dramatic function and a beginning, a middle, and an end.

How to start a conversation 

You don’t need a prompt sheet, but it is good to think about what you expect Claude to do. Here’s an example of I ask Claude to act as a copy editor:

Hi Claude! As a professional copy editor with a keen eye for detail and a deep understanding of language, style, and grammar, your task is to help me refine and improve written content based on these steps:

1. Identify areas that need improvement in terms of grammar, punctuation, spelling, and style.

2. Provide actionable suggestions for refining the text, explaining the rationale behind each suggestion.

3. Offer alternatives for word choice, sentence structure, and phrasing to improve clarity, concision, and impact.

4. Ensure the tone and voice of the writing are consistent and appropriate for the intended audience and purpose without changing the personal style of the author. 

5. Check for logical flow, coherence, and organisation, suggesting improvements where necessary.

6. Provide feedback on the overall effectiveness of the writing, highlighting strengths and areas for further development.

7. Finally at the end, output a fully edited version that considers all your suggestions. 

8. Highlight any changes you made for easy comparison.

The text is a [type of writing, e.g., short story, novel chapter]. The intended audience is [description of target readers].

It’s best to upload a draft that’s already at a stage you would share with a writing group. The more advanced the draft is, the more useful the feedback will be.

If any revision suggests over 20% change or doesn’t sound or feel like my writing anymore, I take a step back and revise the prompt. 

Using Claude as a critique group or beta reader 

For critique group or beta reader feedback, I use fewer steps and do not ask for an edited version.

How to start a conversation 

Here’s an example prompt:

Hi Claude! As a beta reader with a keen eye for detail and a deep understanding of the speculative fiction genre, can you help me refine and improve written content based on these steps:

1. Identify areas you like. Give feedback on characters, setting, structure and pace, explaining the rationale behind each suggestion. 

2. Identify areas you think need more work. Give feedback on characters, setting, structure and pace, explaining the rationale behind each suggestion. 

3. Provide actionable suggestions for improving the text, explaining the rationale behind each suggestion.

4. Check for logical flow, coherence, and organisation, suggesting improvements where necessary.

5. Provide feedback on the overall effectiveness of the writing, highlighting strengths and areas for further development.

The text is a [type of writing, e.g., short story, novel chapter]. The intended audience is [description of target readers].

The feedback and suggestion will be different from prompt 1, and you can switch between prompts or adjust prompts further as you go. Change the persona of your reader to get feedback from different angles.

Conclusions (So Far)

I believe Claude can improve my writing workflow, but it’s crucial to maintain a unique voice and style. If you are new to this, experiment with different prompts and approaches to find what works best for you. AI can’t replace human editors or beta readers, but it can improve the draft before it gets to them.

As with any feedback, human or AI, you are allowed to dismiss suggestions and come up with better solutions—it’s your unique story, only you know what feelings and emotions you want to revoke in your readers.

---

*) the legality on how the companies behind the AI chat bots obtained that data is currently discussed in courts around the globe. It is and will remain questionable in the best of cases.

—

Other news

The first locally printed New Zealand edition of The Fragments is be available on my website. It looks amazing … if I say so myself.