The verdict in three sentences
Buyers of luxury new-build homes in San Francisco, and the brokers who advise them, now ask AI assistants like ChatGPT to shortlist developers, compare projects and check delivery records before they book a sales gallery visit. Whether your firm is named in those answers depends on consistent public facts, credible third-party mentions and project pages an assistant can quote without guessing, and no provider can promise that you will be recommended. The sensible first move is a dated diagnostic: record what assistants say about you today, then decide how much to invest.
What buyers now ask AI assistants about San Francisco developers
The questions have shifted from "luxury condos SF" typed into a search bar to full sentences that read like a broker briefing. The examples below are typical of what move-up buyers, relocating executives and overseas clients ask in 2026.
| Question asked to the assistant | Who typically asks | What the assistant needs to find |
|---|---|---|
| "Which developers are building luxury condos in Pacific Heights or Russian Hill right now?" | Local move-up buyer | One page per current project with neighborhood, status and completion window |
| "Who developed this Mission Bay building and was it delivered on schedule?" | Buyer's broker doing due diligence | A delivered-projects page with dates, plus dated press coverage |
| "New-construction penthouse in SoMa or a restored Victorian in Noe Valley?" | Relocating tech executive | Neutral explainer content that names your projects factually |
| "What HOA dues should I expect in a new luxury building in San Francisco?" | First-time luxury buyer | An FAQ giving ranges clearly labelled as estimates, with a date |
| "Which SF developers work with international buyers and offer remote tours?" | Overseas buyer via a broker | A page on the remote buying process, tours and languages spoken |
| "Are the renderings for this project accurate and when is completion expected?" | Cautious buyer | Renderings labelled as artist impressions, dated construction updates |
| "Who are the architects behind the new towers near Salesforce Park?" | Design-led buyer | Architect credits on each project page and in press material |
When your site is a single launch page with a waitlist form, the assistant fills the gaps with listing portals, old news items and forum threads. That is how a sold-out building ends up described as available, or a project gets attributed to the general contractor instead of you.
Why some developers appear in answers and others do not
Assistants tend to name firms whose public footprint is coherent and easy to verify. The contrast usually looks like this.
| Signal | Developers often named | Developers rarely named |
|---|---|---|
| Firm description | Same name, founding year and focus on the site, LinkedIn and press kit | Different descriptions on every platform |
| Project pages | One lasting page per project: area, architect, status, unit types, timing | Launch pages removed after sell-out |
| Track record | Delivered buildings with completion dates and real photos | Renderings only |
| Third-party mentions | Local business press, architecture media, broker partner sites | Paid placements only |
| Structured data | Organization, project and FAQ markup consistent with visible text | None, or markup that contradicts the page |
| Reviews and testimonials | Genuine, where your brokerage rules allow them | None, or unverifiable quotes |
| Freshness | Construction updates dated every month | Last news item 18 months old |
Content written so that an assistant can quote it is still advertising. Fair housing rules govern how homes, buyers and neighborhoods are described in US real estate marketing, and California consumer protection rules apply to claims about off-plan projects, so a rendering presented as a finished view or an unsupported "best views in the city" line is a risk in any channel. Enquiry forms fall under California privacy law (CCPA as amended by CPRA), and accessibility of the site to the WCAG 2.1 or 2.2 AA level is a sensible target given how frequent ADA Title III website suits are. This is not legal advice: have your counsel review anything that touches these areas.
For reference, Kolonell built the website of Orven Properties, a real estate firm in Dubai. The same principle of one factual, lasting page per project applies, and we do not publish figures from that engagement.
What the work covers and what it costs on the market
In broad terms, the work combines an audit of what assistants currently say, consistency of the firm's public information across the site, directories and press kit, factual project and company pages, structured data, a plan for credible third-party mentions, and a recurring dated record of answers. It overlaps with classic SEO and with public relations, which is why quotes vary so much.
| Item | 2026 ballpark (USD) | Comment |
|---|---|---|
| Initial diagnostic (30 to 60 questions, several runs) | 3,000 to 8,000 one-off | Typical agency quotes in the Bay Area |
| Rework of project and company pages | 8,000 to 30,000 one-off | Depends on the number of projects and photography |
| Digital PR and third-party mentions | 3,000 to 10,000 per month | Public agency rate pages |
| Ongoing monitoring and content updates | 1,500 to 5,000 per month | Often bundled with SEO |
| Full-service retainer, search plus AI visibility | 6,000 to 15,000 per month | Boutique San Francisco agencies |
| Internal time, marketing or sales director | 10 to 20 % of one role | Approvals, facts, photos |
The market is young and these figures are wide on purpose. At Kolonell this work is on quote, scoped after the diagnostic.
How results are measured
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Measurement relies on a dated record: the same set of buyer questions, worded identically, asked at fixed intervals, with the full answers stored. From that record you follow the share of answers that mention the firm, the share that mention each project, the accuracy of what is said (status, architect, completion timing) and which sources the assistant appears to rely on. On the business side, add a "how did you hear about us" field on the enquiry form with an option for AI assistants, and compare quarter to quarter. Answers fluctuate, so trends over three to six months matter more than any single snapshot.
Mini case study
Illustrative example: Bayline Residences Partners (fictional), a San Francisco developer with two delivered buildings and one project in pre-sales, runs a diagnostic on 40 questions. The firm is named in 6 answers out of 40, that is 15 %, and two answers attribute its delivered building to another company.
The year one budget is set as follows: diagnostic 6,000 USD, rework of project pages 18,000 USD, six months of monitoring at 2,500 USD, so 15,000 USD. Total: 6,000 + 18,000 + 15,000 = 39,000 USD. Against a project marketing budget of 1.2 million USD, that is 3.25 %. The team sets a review point at six months: share of mentions, accuracy of the two wrong attributions and enquiries citing an assistant.
FAQ
Can an agency promise that ChatGPT will recommend our project?
No. Assistants compose each answer from many sources and change over time, so any promise of a recommendation is a red flag. What can be measured is the share of answers that mention you, for example moving from 15 % to a target you set, and the accuracy of what is said.
How long before changes show up in answers?
Expect two to six months for corrected facts and new pages to be reflected, depending on how often sources are re-read. Third-party coverage usually takes longer than changes to your own site.
Should we publish pricing and availability?
Publishing a "from" price range and a dated availability status helps assistants answer accurately, while exact unit pricing can stay with your sales team. A range updated at least every 30 days is better than a figure that goes stale.
Does this replace traditional SEO?
No, the two overlap: clear factual pages and credible mentions serve both. Many developers keep 60 to 70 % of the budget on search and test AI visibility with the rest.
What does a first diagnostic cost?
On the San Francisco market, a 2026 ballpark is 3,000 to 8,000 USD for a one-off diagnostic. At Kolonell it is on quote, based on the number of projects and questions.
Find out how AI assistants describe your firm. We record what assistants say today about your company and your projects, then propose a scope matched to your launch calendar. Request a diagnostic quote: reply within 48 hours.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
