Using AI Tools to Research and Write Place

AI can be useful for writing, but only if you give it the right job. Use these tools as research companions, not as substitutes for evidence, judgment, or lived accountability.

The worst use is asking a chatbot to “describe Paris” or “make Tokyo feel real.” That kind of prompt invites the machine to hand you polished fog: attractive phrases, familiar images, confident generalities, and details that may or may not belong together. The result often sounds fluent and hollow.

The better use is to make AI help you ask sharper questions, organize sources, test plausibility, and notice missing pressure.

Think of AI as a research assistant, map partner, scene examiner, and skeptical reader. Keep the authority with yourself. Keep the evidence close. Keep the writing human.

I use three kinds of AI help for place work.

First, regular chat is good for brainstorming and interrogation. I can ask it to help me identify what I need to know before writing a scene. I can ask it to challenge a route, list likely friction points, or help me think through what a character would notice based on class, age, mobility, language, fear, desire, or familiarity with the place.

Here is a useful prompt:

I am writing a scene set in [city/neighborhood/type of place]. My character is trying to [specific goal]. They are [local/visitor/returning after years/new immigrant/commuter/worker]. Help me identify the practical pressures that could affect the scene: route, cost, weather, access, timing, sound, safety, language, etiquette, and what the character might misread. Do not write the scene. Ask me the ten research questions I need to answer first.

That prompt does something important. It stops the AI from writing fake prose too early. It turns the tool into a question generator. Questions are safer than invented authority.

Once you have gathered sources, NotebookLM becomes useful because it works from materials you provide. You can upload or link your notes: articles, city guides, transit pages, neighborhood histories, interviews, PDFs, maps you have summarized, your own observations, and excerpts from books you are studying. Then you ask NotebookLM to help you synthesize from that source set.

Try this:

Based only on the sources in this notebook, create a place dossier for a fiction writer. Focus on details that would affect a character’s behavior in a scene: movement, access, sound, weather, class signals, public/private boundaries, and common mistakes an outsider might make. Separate confirmed source details from possible fictional uses.

That final sentence is crucial. You want the tool to help you separate evidence from application.

NotebookLM is also useful for contradictions. Places are complicated. One source may describe a neighborhood as fashionable. Another may describe long-term residents being displaced. Another may focus on transit access. Another may talk about crime, nightlife, schools, flooding, tourism, or preservation. Instead of forcing one clean summary, ask the tool to preserve tension.

Prompt it this way:

What tensions or contradictions appear across these sources about this place? Which details would matter to different characters: a long-term resident, a visitor, a service worker, a teenager, a wealthy newcomer, an elderly person, or someone returning after ten years away?

That kind of prompt helps your city avoid flattening. It reminds you that no place feels the same to everyone.

NotebookLM’s Audio Overview can also help, especially when you are tired of staring at notes. I would use it as a listening pass. Generate the audio, then listen for ideas that keep returning. Do not treat the audio as final research. Treat it as a way of hearing your source pile discussed out loud. Sometimes hearing the material helps you notice the human problem hidden inside the facts.

ChatGPT Deep Research is useful at an earlier stage, when you need a broad, source-based report. For example, if your story has a scene in a historic market district in a city you have never visited, Deep Research can help gather sources about the district’s history, transit, recent changes, weather vulnerabilities, architecture, and public life.

A strong Deep Research prompt might look like this:

I am writing fiction set in [specific place]. I need a source-based research brief, not travel copy. Focus on how the place works for people moving through it: transit, walking routes, class/access boundaries, weather, soundscape, architecture, public/private space, recent changes, and historical layers. Include citations and links. Flag uncertainty. Avoid stereotypes and tourist clichés. Give me a final section called “Scene pressures a novelist could responsibly use,” with each item tied to a source.

That prompt asks for research, not prose. It also asks the tool to flag uncertainty. You want that. Uncertainty is part of honest research.

After Deep Research gives you a report, do not move straight into drafting. Audit it.

Open the most important sources. Check whether the claims are supported. Look for dates. A neighborhood article from 2014 may no longer describe the present. A transit page may have changed. A blog post may be one person’s experience. A tourism site may smooth over the very friction your scene needs.

Then bring the verified notes into a regular chat and ask for scene pressure.

Use this prompt:

Here are my verified notes about [place]. My scene is about [character] trying to [goal]. Help me choose one place detail that can create pressure in the scene. I want one strong detail, deeply developed, rather than a list. Explain how the detail could change the character’s behavior.

That prompt matches the craft of this course. One strong detail. Deeply developed. Character behavior.

You can also use AI as a cliché detector.

Try this:

Review this scene for place clichés, tourist imagery, vague atmosphere, and details that feel unearned. Do not rewrite it. Point to specific sentences. For each problem, tell me what research question I should answer to make the place more grounded.

That is one of the best uses of AI for craft. It keeps the writing yours. The tool becomes a diagnostic reader.

For a more advanced pass, ask AI to test the scene from different character positions:

Read this scene as if you are three different people: a long-term resident, a first-time visitor, and someone who works in this location every day. What would each person find convincing, missing, or suspicious? Do not invent facts. Focus on questions I should verify.

This helps because place is perspective. A hotel guest, housekeeper, taxi driver, teenager, landlord, nurse, commuter, tourist, and longtime resident do not move through the same building in the same way. AI can help you remember to vary the angle, but you still need sources and human judgment.

Here is another prompt I like for urban scenes:

My character is trying to hide [emotion/desire/fear]. The scene takes place in [place]. Suggest three ways the physical environment could expose what the character is hiding. Keep each suggestion grounded in practical behavior, not symbolism.

This prompt connects place to character pressure. It keeps the output from drifting into decorative mood.

For NotebookLM, build separate notebooks by project or city. A notebook for “Tokyo night research.” A notebook for “New Orleans built memory.” A notebook for “Los Angeles gates and access.” Add your sources. Add your own scene notes. Ask the notebook to create a briefing document, then ask questions against that briefing.

Useful NotebookLM prompts:

• Create a fiction research brief from these sources focused on what a character could see, hear, misunderstand, avoid, or be forced to do.
• Pull out all details related to thresholds: entrances, gates, lobbies, doors, host stands, security, private/public boundaries.
• Based on these sources, what would change between day and night in this location?
• Identify details from the sources that could affect a walking route: distance, hills, crossings, transit stops, weather, crowding, safety, or construction.
• Find details that show historical layering: former uses of buildings, renamed streets, changed businesses, preservation, demolition, or surviving architectural features.
• What should a writer be careful about when representing this place based on these sources?

That last prompt matters. AI can help you avoid overconfidence if you ask it to.

Now let’s talk about using AI to generate draft prose.

Be careful.

AI prose about place often comes out too smooth. It likes atmospheric adjectives. It likes perfect little contrasts. It likes cinematic weather, glowing windows, narrow alleys, distant sirens, and the sort of sentence that sounds wise until you ask what it actually shows. If you use AI to draft, use it for rough possibilities only. Then strip the language back to the character’s actual encounter.

A better workflow is this:

1. Research with sources.
2. Ask AI for questions.
3. Verify details.
4. Choose one scene pressure.
5. Draft the scene yourself.
6. Use AI to critique specificity, plausibility, and cliché.
7. Revise by returning to the character’s body and behavior.

Here is a final prompt for that revision stage:

Read this scene from The City as Character perspective. Identify the single strongest place detail and the single weakest place detail. Tell me where the city changes the character’s behavior and where the city is only decorative. Do not rewrite the scene. Give me a focused revision plan in five sentences.

That is the right posture.

AI can help you see the draft.

It can help you question the place.

It can help you organize the research.

The actual authority of the scene comes from your choices: the one detail you keep, the one behavior you notice, the one interruption you let matter, the one sound or doorway or route that changes the character.

Use the tools to deepen attention.

Then return to the page and write like someone who understands that every place is more than material.

It is someone’s daily life.