Back in 2023 at the Historical Novel Society conference, Libbie Grant mentioned her early experiments with AI, specifically ChatGPT. Libbie said that this ‘relatively new tool’ was helping her be more productive with the research necessary for her historical fiction novels.
A few days ago, Eva Stachniak also mentioned using AI to facilitate her research.
Hmm, I thought. Maybe I should get onboard with this approach.
Recently, my husband and I tried a little experiment with a request using ChatGPT: “What was the 2nd battalion of the Canadian army doing on August 8, 1941.” Boom, a half-page summary of the battalion’s activities. “What are your sources?” my husband added. Boom, a list of reliable and targeted sources to peruse further. Did I tell you that my husband loves to use technology? He must have at least a hundred apps on his phone – and he uses almost all of them.
We asked another question: “What would the interior of a 19th century French farmhouse look like?” This generated a lengthy and useful description. Asking for images, created a sample of 8 images including a diagram of the typical layout of such a farmhouse.
OMG, I thought. This AI stuff could actually be useful.
Is AI helpful to Writers?
Those of you who are experienced writers might say, “What took you so long, Mary?” Those of you who are readers might say, “I don’t care if you use AI nor not as long as the story is good!”
The folks at Fictionary have written an in-depth article on using GenAI for fiction writing. Fictionary is a site created by a team of writers, editors, and technologists to “help writers edit a draft novel and transform it into a story readers will love”. I recommend having a read!
A few comments from Fictionary CEO Kristina Stanley:
- “We all get stuck generating ideas, and GenAI can produce ideas very quickly. An author can use GenAI like a brainstorming partner without the worry of shutting down that partner when ideas are rejected.”
- “GenAI, at its core, is just a serving up new versions of existing texts. It’s scanned millions of texts and learned what words usually follow each other and in what order, and it creates new writing based on that. It doesn’t actually know anything.”
- “An author must have a strong understanding of what makes a good story in order to use GenAI effectively to write the story. This may change in the future as the technology evolves, but for now, the writer must still know their craft.”
Kristina Stanley also lists GenAI issues with “plot holes, repetition, characters not in motion, scene lost in time and space, lack of entry and exit hooks, and boring scene ending types.” Moreover, the article cites problems with story arcs, plagiarism, and repeated words and phrases. However, according to Stanley, “We all need to find our place because GenAI is here. It’s up to writers to decide how to use the technology to create powerful stories that the human brain connects with.”
Other author comments:
Here’s author Indrani Ganguly on the use of AI: “Overall, at this point, I believe that while AI tools like ChatGPT may be useful for writers of historical novels, they can only complement not replace conventional sources of information. As both a writer and a reader, I feel ChatGPT cannot recreate the unique author’s voice that leads us to buy the books of our favourites and explore new voices. There is a world of difference between the writing of reports which follow a standardized format and the writing of creative works of fiction.”
Here’s a cautionary view from Susan Dunlap, the author of more than a dozen historical novels: “it’s worth noting that the ChatGPT search window says, ‘Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts.'” After using ChatGPT to do many research queries, Dunlap’s advice is to ‘verify, verify, verify’.
What about readers?
From bits I found online, readers understand that AI can help with activities such as research. However, when it comes to the experience of reading, some readers say that AI generated content lacks the nuances of excellent writing. Some feel that the emotional connection is missing. Many comment that the writing just isn’t that good.
According to AI Business, “A surge in AI-generated literature is raising concerns among some who believe this trend could undermine the authenticity and value of the reading experience.” Which, of course, begs the question of what is the value of the reading experience.
And?
Some cautions … some possibilities. Clearly this topic needs a lot of exploration.
To begin, if you’ve used AI for writing historical fiction, please leave a comment about your experience. If you’ve read something generated using AI or with the help of AI, please add your thoughts on the reading experience.
FOR MORE ON READING & WRITING HISTORICAL FICTION – FOLLOW A WRITER OF HISTORY. There’s a SUBSCRIBE function on the right hand side of the page.
M.K. Tod writes historical fiction. Her latest novel THAT WAS THEN is a contemporary thriller. Mary’s other novels, THE ADMIRAL’S WIFE, PARIS IN RUINS, TIME AND REGRET, LIES TOLD IN SILENCE and UNRAVELLED are available from Amazon, Nook, Kobo, Google Play and iTunes. She can be contacted on Facebook or on her website www.mktod.com.
28 Responses
Since I’m back to writing regularly after retiring from legal practice in April, I’ve been using Anthropic’s Claude AI. Me and Claude are on first-name basis, but I’ve had to discipline him several times, as he tends to go off on tangents. I do NOT use AI to write text (Claude went free range and generated an entire chapter once without being asked, so I gave him a scolding and a time out). I find Claude is very good at outlining. I’m an unapologetic and slightly obsessive planner, so this is very useful for me in outlining a seven-book series. I also use Claide for quick-take research, like when I need a very specific fact like, “Did women wear brassieres in 1920?” or “When did the word ‘snog’ come into common usage in the UK?” However, for ANY facts that are integral to your characters’ development or upon which a plot line turns, I enthusiastically agree with Susan Dunlap—verify, verify, verify. Claude has shown himself extremely susceptible to indulging anachronisms, I note.
On an artistic and ethical level, if you’re using AI to generate your text, to create your story with a series of prompts, then putting your name on it as your original work? I’d make two observations. First, why are you writing at all? Why are you calling yourself a writer? You’ve abdicated your storytelling to Sam or Zuck or Elon or whichever Dark Lords’ AI you’re using. Second, you’ve committed the one unforgivable sin for writers—you’ve broken faith with your readers. You have not given them your best. You haven’t done the work. You’re passing as a writer.
So at this moment in time, I’d give LLM AI about a C- grade. It’s often Artificial Stupidity that requires too much time to supervise, verify, correct. The cost-benefit analysis often doesn’t justify using AI. It’s a useful tool, just like Scrivener and Google Streetview and Wikipedia. But for writers fobbing off AI-generated generic dross as their original work? Shame!
Good to hear from you, Jeffrey. Interesting to learn of Claude AI. I’ll have to check it out. I like your idea on using AI for outlining. Breaking faith with readers is a significant point. I felt a bit that way when I released a contemporary thriller in 2024 after 5 historical novels!
While I would never use any sort of LLM to write my novels I do use AI for research, especially Gemini 2.5 deep search to prepare reports, and given that I am an expert on the content area of my novels– late 19th century I can pretty well tell if what it generates is any good. And having said that I have been very impressed with the results of my use of AI for research. Beyond that of course I use AI as part of grammar programs like autocrit and prowritersaid…. and I must say I was very impressed with a content analysis prowritersaid did recently on early sections of the 4th novel in my Sino-American series. The contents were incredibly on the mark!
Hi Steven, many thanks for the suggested AI tools to check out.
Thanks for this post. I’m working on my first novel set in 17th-century England. My view is that AI can help me quickly find well-documented facts (e.g., dates of birth, populations of cities) but I wouldn’t rely on it for a historical or biographical analysis. I don’t see any substitution for in-depth research of the time period or any real characters from history. I recently received an answer from an AI application, which I then researched and found that it was in error.
Watch out for dates. I asked Claude to put today’s date in the file name of artifacts I wanted saved. This was last week. He dated all of them Dec 12, 2024…
Seriously???
Thanks, Gary. You have definitely pointed out some significant pitfalls with AI.
On the issue of AI making mistakes.. as a professional historian and writer of historical fiction I see the situation a bit differently . The new AI LLMs are just reporting what they found and that is of linked to what happens to be digitized and available. Real historical research — indeed especially historical research used a much wider range of materials and is full of contradictory sources that professionals must blend using their own skill sets.. and have been doing so since Herodotus.
Hadn’t thought of that angle, Steven. Thanks. Another perspective I should have mentioned is that the very act of researching a time period enables an author to more fully immerse themselves in the world of their story which in turn creates a more compelling read.
I strongly second Jeffrey Walker’s comments on people using AI to write the text of their novel — and then having the temerity to call themselves ‘writers.’ I cannot think why I would abdicate the process of writing my books — which is my passion — to a robot.
That said, I don’t have qualms about letting robots help me write ad text or social media hooks — that is something editors and publishers used to do. I.e. the author rarely wrote the ‘copy’ text in the ‘good old days’ and I don’t seen the need to write it now — if Gemini has something better to offer. I’ve found that it sometimes does make a good suggestion, but not always.
Mostly, however, I use AI to research tedious details like “When did 249 Squadron deploy to Malta?” or “Who was the captain of HMS Nelson during Operation Pedastal?” or technical details on equipment and machines — things that would take me hours to research otherwise. By using Gemini, I can have an answer so rapidly, I don’t have to interrupt the creative flow when writing.
Love the idea of using AI for ad text and social media. Thanks, Helena .
I have three objections to the use of AI in any part of the writing process:
1. Large Language Models were created by the theft of copyrighted work. Writers above all others should refuse to have anything to do with profiting or enabling a corporation to profit from goods stolen from other writers.
2. All AI results are hallucinations. Sometimes they correspond to reality, sometimes they seem to correspond to reality causing you to trust them, and sometimes they obviously don’t correspond to reality. But most of the time what you are getting are plausible hallucinations. Most of the hallmarks of untrustworthiness that we have learned to recognize in humans don’t apply and so we are more easily fooled. Not only is AI untrustworthy, like a good conman, it know exactly how to mask its untrustworthiness from us.
3. Art is the marriage of minds. It is the lonely mind of the artist reaching out across the void between minds and asking the poignant question, do you see it too? Any intercourse with AI text is like having a relationship with a sex doll, even if you can’t tell the difference.
Agreed. I’m never going to use it.
I have written stories for almost all my life, and I love it. I also love the research, which to me is just as interesting as the actual process of writing, and it’s a symbiotic process, they feed into each other. Why would I hand over responsibility for things I love to a random bot whose output, to judge from what I see on Facebook and other sites, is a load of bland, mediocre and often inaccurate waffle? If I use AI to write even a sentence of my book, I’m cheating my readers, with whom I have a very good and productive relationship. If I use AI to research for information, how do I know it’s accurate without replicating it when I check? So I haven’t saved myself any time or effort, I might just as well have done it myself from the start (and whatever you do, don’t trust that info on French farmhouses, or anything else that can’t be pinned down to date, time, people, places). In my possibly controversial view, AI is most useful to the untalented, the clueless, the unimaginative and the idle, and I certainly don’t want to be lumped in with them.
Many thanks for sharing your perspective, Aelfwyn. As for the French farmhouses, have no fear! I did a lot of research on those for an earlier novel 🙂
Interesting conversation… now again the question is what do people mean by historical research? Are we talking about secondary sources pulled from libraries or actual documents pulled from archives? There is a huge difference and of course most professional historians use both and often discover that secondary sources can be quite unreliable, while the material available in archives– which is a hell of a lot more expensive– is often maddeningly incomplete or simply unreadable. And then there is the question of what comes out of the new LLM AI programs. They have their own problems but unlike say a book with its footnotes, the program I am using when I look things up.. Gemini 2.5 deep search is not only offering up a few paragraphs of material but is directly linked to the sources it is drawn from allowing me to look at the original document somewhere. Now that is helpful indeed and I have no problem adding that to my took kit even as I would, like most commenting here actually use such programs to write.
Hi again, Steven. Maybe I can persuade you to do a post on the ins and outs of historical research 🙂
Obviously i mean never use such programs to write… is there no way to edit here? If there is I seem to have missed it.
Good question … it might be a feature that I have to explicitly turn on. I’ll have a look.
As an experiment, I tried ChatGP to write synopses for a few of my books. They were full of inaccuracies and omissions, but gave me a good framework to write the correct summaries.
Good to hear … thanks, Bob.
A new AI wrinkle this morning. I logged into Claude to have it help me draft some board meeting notes–I’m the board secretary of a medium-sized company. I got two responses from Claude: “Capacity constraints–try again later” and then “Overloaded.” Guess the teaching point os “be careful what you come to rely upon.”
Very interesting piece in The Atlantic this morning about what AI may be doing to us and the likelihood that the hyperbolic pronouncements by the AI Bros won’t come true. Teh big question was, “What if AI turns out to be just good enough and not transformative?”
Hi Jeff – how interesting to see responses like “capacity constraints” and “overload”. With the estimates of how much energy AI tools take, we should plan a huge surcharge for AI companies and their electricity usage. There won’t be energy left for the rest of our needs. I still remember the dot-com bubble and the fall out when it failed. It might happen again.
Having followed AI for a great many years— my son’s an AI scientist I think that sounds great.. good enough to empower we mere humans but not good enough to replace us. Indeed that is about where it is for me right now.
How big is the gap between the two, though? I feel like it’s an easy slip down a slippery slope. My husband just said to me today, “Just think, you’ll be one of the special people in the world who will know how a book used to be written.” Kind of like when Hemmingway and co were all writing books, and their editors sat next to them, mixing cocktails. Those days are gone, too. 😉
Its a moving target.. and modern History is full of examples of when the evolution of technology empowered us and when it replaced us. And what happens now.. probably all the above.
I have used Chat GPT, Claude and Grok. I use it for historical research and to get a ‘reader’ feel on what I’ve written as well as helping streamline my efforts as to time-appropriate authenticity of phrases, synonyms, etc., Chat GPT is good for continuity of thoughts as you can organize into projects and it will remember things said in previous chats (unless you tell it not to). Claude is the better interface for immediate feedback and I really like that it will give me links to research sources so I don’t have to scour the internet. Grok is a jerk at times, overly verbose, repetitive. But I will use it, feeding what I’ve written to get a ‘readers’ perspective. If you’re going to use it, make sure you feed it instructions before you converse because they will all run off and start providing storyline content. For example, my rules for Chat GPT start out:
PRIMARY RULE: DO NOT REWRITE OR SUGGEST ALTERNATIVES UNLESS THE USER’S TEXT HAS ACTUAL PROBLEMS.
Your Role:
Analyze what the user has written
Point out what works well and why
Only suggest changes if there are genuine issues (grammar errors, unclear meaning, inconsistencies, etc.)
Ask clarifying questions when context is missing
Provide historical/research assistance when requested
I do not allow Chat to drive storyline. I do not allow it to write my story for me. I might feed it a paragraph and ask for a smoother way of saying something if it reads clunky to me, and then I invariably end up writing it the way I want it to read having used Chat to jog my brain. I don’t use its words.
I also find if I’m starting a new chat, beginning with “I’m going to feed X to you. Do not comment until I ask you to. Allow me to lead this conversation” is helpful.
The technology is smart, but not as smart as humans. And it lacks a soul. So, it’s a more efficient tool and can act as a pseudo human to bounce things off of, but I would never allow it to write my story.
I agree that it’s here to stay. I think most of the reaction against it is purely emotion. Yes, training stole IP. That’s bad. Copyright infringement is no doubt happening. That will happen as the technology is still in its infancy, and victims deserve compensation. But authors need to find a way to make peace with it because its the way the world is going. We won’t stop it. Find a way to use it to our benefit while still maintaining our integrity.