Gains:
- Being able to distinguish where artificial intelligence saves real time in the real estate workflow (portfolio, marketing, valuation, legal, sales) and where responsibility-critical decisions (final price, legal validity, customer declaration) are left to the expert and the human, according to the task risk level.
- Ability to apply a discipline that verifies each AI output by connecting it to the source, comparing it with your own data, and passing it through a field/legal filter.
- Ability to acquire the habit of anonymizing customer and property data within the scope of KVKK, choosing safe vehicles and avoiding misrepresentation
Hundreds of decisions are made silently every day in a real estate office. What should be the price of an apartment? Search for this ad no matter how it is written; which property suits this buyer; Is there a risky clause in this rental agreement? Does this investment really bring profit? Some of these decisions are repetitive and time-consuming: writing ad text, editing photos, creating market summaries, responding to messages. Others are decisions that directly affect a person's money, legal situation and trust, and where the margin of error should be small: the final price, the statement given to the customer, the validity of a contract. Artificial intelligence (AI, or AI for short—computer systems that can generate text, recognize patterns, create images, and perform calculations like humans) fits right in the middle of this picture: when used correctly, it speeds up repetitive work and frees up time for you to take care of the customer; When used incorrectly, it can carry seemingly safe but false information into an advertisement or customer declaration.
The first unit of this module is not a software introduction. Its purpose is to clarify where to put AI in your profession and where not to put it at all. Because real estate is a "liability-critical" field: every statement you make can affect a buying-selling decision, a loan application, or a legal dispute. Let's put the basic principle from the beginning: Artificial intelligence is an assistant, not a consultant or expert. Responsibility for the final price, legal validity and declaration given to the customer belongs to the competent expert - licensed real estate consultant, sworn appraiser, lawyer. ("Authorization Certificate" is the Real Estate Trade Authorization Certificate; it is the official document required to trade real estate.)
Map of the real estate process and the place of AI
To understand the business, let's break down the real estate process into five areas. The portfolio and data stage is where the properties of the property are collected and recorded. The marketing phase consists of ad text, visual, video and promotional content. The valuation phase is determining the price. The sales/lease phase is customer matching, showing and negotiation. The legal and closing stages are the contract, title deed and transfer transactions. AI can touch every stage, but not with the same authority. You can largely have AI write marketing copy; But you can't make the AI "decide" the legal status of a title deed or the exact value of a property.
Let's define a few basic terms from the beginning. AVM (Automated Valuation Model) is a system that produces an automatic price estimate for a property by using historical sales data; It gives an indication and does not replace official expertise. A comparable is a recently sold/rented comparable property that is similar to the property being appraised. Appraisal (valuation report) is the official report prepared by an authorized/sworn appraiser by conducting an inspection in the field. KVKK is the Personal Data Protection Law; It protects data such as customer name, phone number, identity and title deed information. A prompt is a written instruction you give to the AI. Since these terms will recur throughout the module, we ground them here.
The following table summarizes the role and risk level of AI by mission:
Quest
Role of AI
Risk level
Who approves
Advertisement text / description draft
sketch generator
Low-Medium
Consultant (confirms information)
Photo enhancement, virtual staging
visual assistant
medium
Consultant (non-misleading limit)
Market summary / region analysis
compiler, interpreter
medium
Advisor (verifies with current data)
Value estimate (AVM)
indicator forecast
Medium-High
Consultant / expert if necessary
Contract risk screening
Pre-browser
high
Lawyer (legal opinion)
Final price/declaration to customer
not helpful
high
Consultant (signature and responsibility)
Keep in mind the one line in this chart: as risk rises, AI's role shrinks, human approval grows.
Why "verification" is the heart of this business
Artificial intelligence language models seem confident in their answer, but they may not be sure. In technical language, this is called hallucination: it is the model's fabrication of non-existent information in a fluent sentence, as if it were true. This is a serious trap for a real estate consultant: the model may tell you "the cost per square meter in this region is around 45,000 TL" (maybe it is close to the truth), but it may also make up a zoning situation that does not actually exist, saying "there is a 4-storey development right on this parcel". Since he says both with the same fluency, the only thing that distinguishes right from wrong is your field knowledge and verification habit.
The verification discipline consists of three steps:
- Link to the source: For objective information such as zoning status, m², license, title deed record, dues, not to the AI's memory; Rely on the land registry, municipality/zoning information, license, management plan and your own measurement. Use AI to comment on that data, not to remember it.
- Recalculate / compare: Independently check each numerical result (price, yield, m² unit price) given by the AI. In particular, check the currentness of the precedents and their compatibility with the property (floor, facade, age).
- Field and legal filter: Evaluate for yourself whether the output conflicts with the actual status of the property, zoning/legal status, and customer needs.
Attention: Telling the customer about the information produced by the AI without verifying it is like accepting an unsigned document as official. Just because the information comes out fluent is not true; It is correct if the source can be shown.
Privacy: customer and property data is personal data
Customer information (name, telephone, TR ID, income status, credit information) and identifying data of the property (full address, title deed block/parcel, owner name) are personal data within the scope of KVKK. Pasting a customer's name, phone number and title information onto a public AI tool is a serious violation. The rule is simple: anonymize the data. Instead of "Ahmet Demir, 0532..., Bağdat Cad. No 45 D8, title deed island 1234 parcel 5", write "3+1, 130 m², 4th floor flat in Kadıköy, sea front"; Remove name, contact, full address and title information. If possible, choose corporate tools that have a data processing agreement and do not use your data in model training.
Four copyable starter templates
You can use the four templates below starting today. The rule common to all of them: no identification data, fabrication prohibited, missing information is marked with square brackets.
Template 1 — Anonymization control:
Is there personal/identifying data (name, phone number, TR ID, full address, title deed/parcel, e-mail) in the text below? If so, list them and suggest how to anonymize each one. Text: [paste text]
Template 2 — Validation list generator:
Role: You are a meticulous real estate consultant. Task: Remove each OBJECTIVE information (m², room, floor, license, dues, heating, age) in the ad draft below and make a table with a "from which source should it be verified" column for each. fabricating information; list only those mentioned in the text.Text: [paste draft]
Template 3 — Draft polite response to customer message:
Role: You are a professional, warm but understated real estate agent. Task: Draft a brief (60-90 words) response to the client question below. Do not use warranty/certainty; Say "I will confirm and get back to you" regarding price and legal issues. Question: [paste anonymous question]
Template 4 — Task risk classifier:
I will perform the following task: [write task]. Classify this as low / medium / high risk and itemize which steps require human approval and which information needs to be verified from the source.
three mini cases
Case 1 — Safe use. A consultant requests a draft advertisement text for the 3+1 flat he has. Prints draft to AI with anonymous attributes (no name, full address). Then he confirms the m² with the value in the title deed, the dues with the management plan, and the heating type with what he sees on site; pulls out the over-the-top “opportunity of a lifetime” line. Duration: 5 minutes instead of 20 minutes. AI gave the draft, the responsibility remained with the consultant.
Case 2 — Unverified information trap. Another consultant asked YZ, "What is the average dues for apartments in a complex in this region?" he asks; YZ says "about 1,500 TL". The consultant writes this in the ad. However, the real fee for that site is 3,200 TL. When the buyer learns this at the first screening, trust is shaken and a misrepresentation controversy arises. The right way was to confirm the dues from the management.
Case 3 — Breach of confidentiality. An assistant loads the customer's full name, phone number, and title photo into a public AI tool and "sums it up." The data went to an external server. The office faces a KVKK risk. The right way: to share only necessary, de-identified information.
Weak prompt / Strong prompt
Weak prompt:
Ahmet Demir's Bağdat Cad. Write an ad for his flat at No 45 D8, call 0532... You decide the price.
This request is wrong in three aspects: identity and contact information were shared (KVKK violation), it left a responsibility-critical decision such as price to the model, and no verified features were given. AI fills in the gaps with prediction.
Powerful prompt:
Role: You are an experienced real estate consultant. Task: Write a draft of the advertisement description for sale with the following VERIFIED features. Property: 3+1, 130 m² gross, 4th floor (with elevator), 12 years old, natural gas combi boiler, within the site, sea frontage, Kadıköy. Rules: Use only the information I have given, the information is FITTING, if there is missing information, write [to be confirmed] in square brackets. Do not use guarantees/exaggerations. Length: 120-150 words. Write a price.
The difference: there is no identification data, the price decision is in the hands of the consultant, he is directed not to fit the model but to leave it incomplete.
Common mistakes
- Using information made up by AI without confirming it (zoning, dues, license). The most common and risky mistake.
- Let the price be determined entirely by the model. Price is a responsibility-critical decision; The model gives an indication, you decide.
- Pasting customer/property ID data into the public tool. KVKK violation.
- Putting the responsibility on the vehicle because "AI said so". You are responsible for the declaration going to the customer.
- Mistaking fluent output as correct. Just because it's well written doesn't mean it's true.
Tip: For the AI output going to each customer, ask yourself: “From what source can I point this information?” If you don't have a response, that information can't be published yet.
In summary
Artificial intelligence is a powerful assistant in real estate: it saves serious time in the production of listings, marketing, market briefs and drafts. But the decision is up to the person. As the risk increases (price, law, statement), the role of AI becomes smaller and your approval grows. Reinvent three reflexes: anonymize identity data, attribute any objective information to the source, trust verified output, not fluid output.
Application task
For a property you hold: (1) prepare an anonymous property listing, removing all identification/contact/title information. (2) Have a draft ad produced using the "Powerful prompt" template above. (3) With Template 2, connect each objective information (m², dues, license, floor) to a source and mark it as "confirmed / not confirmed". (4) Verify that you have removed the price phrase from the text.
checklist
- [ ] I anonymized customer and property data (no name, telephone, full address, title deed information).
- [ ] I linked every objective information to a source (deed, license, management plan, measurement).
- [ ] I did not leave decisions such as price and legal validity to the model.
- [ ] I have removed exaggerations and guarantees.
- [ ] "From what source can I show this information?" I asked the question for each output.