Why AI-written marketing reads as generic, and what fixes it

A general-purpose language model has never walked past the hardware store on Third Street, never noticed when the bakery changed hands, and never stood in the rain waiting for the farmers market to open at eight in the morning. It cannot name what it has not seen. That limitation shows up the moment you ask it to write about a place: the prose sounds plausible, but the details stay vague. The event is mentioned but not dated. The restaurant is recommended but not located. The street is admired but never named.

Local marketing works the opposite way. It assumes the reader already lives in the neighbourhood, or sold a home there once, or is trying to picture what it would be like to raise a child three blocks from the park. The value is not in the idea of community but in the fact that the library added Sunday hours last month and the new playground opens this Saturday at ten. Generic copy violates the first rule of that contract: it cannot be forwarded to someone who might move back, because it does not say anything a search engine could not.

A sentence that names Maple Avenue and the corner where it meets Tenth is automatically more useful than a sentence about tree-lined streets and walkable blocks. The first gives a reader something to look up. The second gives them a phrase they have read in a dozen other emails. The difference is not tone or style. It is whether the writer has committed to a specific place and left evidence of that commitment in the text.

This does not mean every sentence needs an address. It means that when you mention a business, you name it. When you mention an event, you say when it happens and where it starts. When you talk about a park, you use the name the city gave it, not a general category. The reader may not go, but they can tell you know the answer.

An AI prompt that asks for farmers market content will return sentences about fresh produce, local vendors and weekend mornings. A useful market update says the market runs Saturdays from eight to noon through October and starts again in May. One version fills space. The other helps someone plan a Saturday.

The same principle applies to new businesses, public meetings and seasonal programs. If the library is hosting a lecture series, the newsletter should say it starts on the third Tuesday of the month at six thirty. If a restaurant opened last week, the newsletter should say it serves dinner Wednesday through Sunday starting at five. Specificity is not about sounding informed. It is about being correct.

A template can describe spring markets or fall festivals in general terms, but it cannot tell you that the art walk is this Thursday or that the planning commission meets Tuesday night to vote on the zoning change everyone has been arguing about. Those details require someone to check a calendar, read the city agenda or drive past the banner going up outside the gallery. They also make the newsletter worth opening, because they answer a question the reader did not know to ask yet.

This is where the weekly rhythm matters. A letter that goes out the same day each week can carry news that is still current when it arrives. A letter written once and reused cannot, because the event it mentions has already happened or the deadline has already passed. The fix is not better writing. It is a commitment to research the week you are writing about, not the season.

Voice is not decoration. It is the set of choices that makes one agent sound like themselves and another sound like someone else. Those choices include sentence length, whether you use contractions, how often you editorialize and whether you open with the news or with a personal note. A voice chosen once and applied consistently becomes recognizable. A voice that changes every week because a different template was used becomes invisible.

The way to test whether a voice is yours is to read the draft out loud and notice where you would say it differently. If the sentence sounds like something you would write in a text to a friend who just moved to town, it probably fits. If it sounds like the opening paragraph of a brochure, it probably does not. The goal is not to be casual or formal. It is to be consistent enough that someone who reads three issues in a row can tell they came from the same person.

A tool that generates content without input will generate the same kind of content for everyone using it. A tool that asks the agent which topics matter and which voice to write them in can produce something that reflects a choice. The difference is whether the agent is approving a template or shaping the result.

This is also where review matters. If the draft arrives a day before it sends and the agent can change any part of it, the tool is working for the agent. If the draft goes out automatically with no chance to edit, the tool is working instead of the agent. The first approach treats writing as a collaboration. The second treats it as a replacement. The collaboration version produces better letters, because the agent can add the detail the system missed or remove the section that does not fit the week.

An agent in one city will care about the school board meeting and the rezoning proposal. An agent two towns over will care about the new traffic pattern and the park expansion. A tool that writes the same housing update for both of them is not writing about either market. It is writing about the idea of a market, and the idea is not what keeps someone reading.

The fix is to write about the boundaries you work in, using the names the residents use. If you cover three neighbourhoods, name them. If you cover a whole county, say which parts you know best. If a street is the dividing line between two school districts, mention it. The reader does not need a geography lesson, but they do need to know whether the information applies to them. A newsletter that names its territory earns the right to be specific about it.

Even a well-trained system cannot tell you that the Italian place on Fourth Street started serving brunch last Sunday, or that the city just posted the budget hearing date, or that the house on the corner finally sold after sitting for eight months. Those details come from driving the neighbourhood, reading the local paper and paying attention to what changed since last week. They also make the difference between a newsletter that could have been written by anyone and a newsletter that could only have been written by you.

This does not mean you write the whole thing from scratch. It means you add the sentence the tool could not, and you remove the sentence that does not fit. The result is a letter that sounds like it came from someone who lives in the market, because it did.

Local Letter is a weekly local newsletter service for real estate agents and lenders. Each week it researches and writes a newsletter about the agent's market, including local events, housing updates and places worth knowing, in the agent's voice and under the agent's brand.

The agent chooses from eight voice styles and can blend up to three of them. There are twelve section types to choose from, and Local Letter rotates among the sections the agent selects, so the email keeps moving without being rebuilt. Each issue can carry a written feature story drawn from fourteen subjects the agent chooses, and the agent can remove the story from any week's issue in the review.

About a day before the send slot, the issue is emailed to the agent to review. The agent can approve from the email itself, change any piece in the app, write the note in their own words, or skip the week. An issue that is not approved expires and never sends.

Every plan starts with fourteen free days, during which the letter can reach up to ten people, and the agent can cancel anytime.