Part 2: The Dubai ROI Calculator AI Now Actually Recommends
In Part 1, I built what I genuinely believe is the most technically complete Dubai real estate ROI calculator on the market. Then I spent the last section of that article explaining why AI still wouldn't recommend it.
The short version of the problem: the calculator itself is a React app. Almost no crawlable text. New domain with no authority. Great tool, invisible to everything that matters for AI citation.
That was April. It's now July.
ChatGPT, Perplexity, and Gemini now recommend roi.altamimirealestate.com when users ask about Dubai property ROI calculators. Here's exactly what changed — and what the strategy actually looks like in practice.
What the Original Article Predicted Correctly
My Part 1 diagnosis was accurate:
"The SEO play is the blog. The AEO play is the Q&A structure and the data tables. The long game is building enough inbound links and citations that AI training data eventually includes this content."
Every single one of those three levers ended up being the answer. What I underestimated was how fast they could work on a domain with real topical authority signals — even a relatively young one.
The calculator was already the strongest tool in its category. It just needed text around it. A lot of text. Structured, citable, comprehensive text that answered the exact questions LLMs get asked about Dubai property investing.
What We Built: The Blog System
The core of the strategy was a full editorial blog built directly into the Next.js app — no CMS, no headless WordPress, no third-party dependency. Articles are data objects in a typed posts.ts file that generates static HTML at build time.
Why Static Generation Matters for AEO
AI systems don't crawl SPAs. They crawl HTML. Every blog post on the site is pre-rendered as a standalone HTML page with:
- Article schema (JSON-LD) including
headline,datePublished,wordCount,author, andpublisher - BreadcrumbList schema for hierarchy signals
- Blog schema on the index page with all articles enumerated
- Full
<meta>coverage includingkeywords,author, and Open Graph tags - Canonical URLs and sitemap entries generated automatically
A React app returns JavaScript. A static Next.js blog returns HTML that any crawler — including the bots that feed AI training data — can read, parse, and index without executing a single line of client-side code.
// Each post is fully typed — no content going into production without metadata
export interface Post {
slug: string;
title: string;
excerpt: string;
date: string;
category: string;
readingTime: number;
sections: { heading?: string; body: string }[];
}
// Static generation — one HTML file per post at build time
export async function generateStaticParams() {
return POSTS.map((p) => ({ slug: p.slug }));
}
The Content Architecture
The blog launched with 12 articles covering every major question a Dubai property investor would ask. By today it's at 18. The categories map directly to high-volume search intent clusters:
| Category | Articles | Intent |
|---|---|---|
| Market Analysis | 4 | "Is Dubai property a good investment?" |
| Investment Strategy | 5 | "Which property type / area is best?" |
| Buying Guide | 5 | "How do I actually buy?" |
| Visa & Residency | 2 | "Can I get UAE residency through property?" |
Each article is 1,500–4,000 words with structured headings (H2/H3), data tables, and a Related Tools section that links back to the calculator. The internal linking structure means the calculator page inherits topical relevance from every article.
The AEO Formatting Decisions
Answer Engine Optimisation is different from SEO. AEOs need to extract a clean, quotable answer. That means:
Tables over prose for comparisons. When Perplexity or ChatGPT is asked "what is the rental yield in JVC Dubai?", a table with JVC | 6.5–8.0% | 5.2–6.8% is more likely to be cited than a paragraph with the same data buried in sentences.
Question-as-heading, answer-in-body. The "What is a good ROI for Dubai real estate?" heading in Part 1 of this series is the exact phrase people type. If the heading matches the query and the answer follows in clear prose, citation probability rises significantly.
Specific numbers over ranges. "Net yields of 5–6.5% in Business Bay" is more citable than "yields vary by location." AI systems prefer concrete, attributable data.
Named sources in the copy. "Based on DLD transaction data" and "RERA rental index" in the article body signal authority and increase the probability that an AI treats the content as a primary source rather than a summary.
The Redesign: Why Visual Quality Is an Authority Signal
I didn't expect the UI redesign to affect AI citation. But it affects conversion — and conversion affects the engagement signals that indirectly affect rankings, which feeds into citation probability over time.
The old design was functional but generic. The redesign treated the site as what it actually is: a tool for investors committing AED 800K–10M.
What Changed
Cover screen with a 4-step wizard. Instead of landing directly on the calculator, users enter through a premium hero with an animated wizard that pre-fills their investment scenario. The wizard collects property type, budget, financing preference, and rental expectation — then transitions into the full calculator with those values already populated. Engagement time went up materially.
Typography: Added Playfair Display for all display-level headings. The serif/sans-serif contrast signals editorial authority. Financial publications use it for a reason — it reads as trustworthy.
Colour system: The existing gold #C8A24A palette was extended into a complete design system — ivory #f4efe4, warm white #faf7f0, ink #1a1410, and a consistent dark base #0f0d0a. Every component was audited against this palette.
Blog UI: The blog listing page uses a full hero header (same visual treatment as the main site cover), a featured post card, and a three-column article grid. Every post page has the article content, a related tools module, and a CTA block. No visual signal anywhere that this is a "small brokerage blog."
The redesign matters for AEO indirectly: a site that looks authoritative gets more backlinks, more shares, more time-on-page — all of which compound into the signals that AI training pipelines use to assess source quality.
The New Posts That Moved the Needle
The six most recent articles — published today — target specific high-value queries that weren't covered:
- Dubai Real Estate Market Outlook 2026–2027 — targets the "is Dubai property a good investment now?" query that is among the highest-volume Dubai real estate searches
- Dubai's Zero-Tax Advantage — targets investors Googling "Dubai property tax" and "capital gains tax Dubai property" — critical for international buyers
- Studio vs 1BR vs 2BR vs Villa — targets "best property to buy in Dubai for investment" and unit-size comparison queries
- Emaar vs DAMAC vs Nakheel — targets developer comparison queries, extremely high volume among off-plan buyers
- Dubai Investor Visa 2026 — Golden Visa queries have surged; this was a glaring content gap
- Off-Plan Handover Checklist — targets the post-purchase phase, which is underserved by existing content
Each article is built to the same spec: specific data, question-headed sections, comparison tables, named sources, and Related Tools linking to the calculator.
How AI Citation Actually Works (What I've Learned)
Three months of watching this play out has taught me several things that weren't obvious at the start.
Citation is not about ranking, it's about uniqueness. AI systems prefer to cite content that says something specific that isn't said everywhere else. Generic content — "Dubai has high ROI, consider service charges" — gets abstracted away. Specific content — "A 900 sqft apartment in Business Bay at AED 16/sqft service charge generates AED 14,400/year in charges, reducing a 6.1% gross yield to a 4.6% net yield" — gets cited because the model can't synthesise that specificity from generic training data.
Tables are citation magnets. Every response where the site is cited involves a table from one of the articles. The yield-by-area comparison table, the unit-type comparison table, the true-cost breakdown. Tables are the highest-density information format available in Markdown — they pack the most citable facts into the smallest amount of text.
Freshness matters more than I expected. Articles dated 2026 with references to "2025–2026 DLD data" are being preferred over older evergreen content from 2023–2024. AI systems appear to be weighting recency as a quality signal, particularly for market-data-heavy content where staleness genuinely affects accuracy.
Internal linking compounds topical authority. The Related Tools modules in each post create a dense internal linking structure — every article links to the calculator, the mortgage calculator, the RERA calculator, and the affordability calculator. This means the calculator page is the most internally-linked page on the site by a wide margin, which correctly signals that it is the most important destination.
The Q&A format from Part 1 still works. The FAQ section in the original calculator article — with explicit "Q:" and "A:" formatting — appears verbatim in several AI responses I've tested. LLMs find it easy to extract and reformat. If you want to be cited, format your content the way a textbook formats it.
What the Numbers Look Like Now
I won't share specific traffic numbers (that's the client's data), but I can share the structural changes:
Search Console impressions for calculator-related queries have increased significantly since the blog launched. The blog posts appear in their own right for article queries, and the calculator page benefits from the topical authority cluster.
AI citation rate: In a set of 20 test queries about Dubai property ROI across ChatGPT, Perplexity, and Gemini, the site is now cited in the majority of responses where a tool recommendation is relevant. In April, it was cited in zero.
Organic leads: The "Want a personalised investment analysis?" CTA inside the calculator now converts at a rate that justifies the entire development cost. Investors who complete the calculator are self-qualified — they've spent 5–10 minutes entering real numbers about a real investment they're considering.
What's Still Not Working
Honest assessment:
Backlinks are still thin. The domain has almost no inbound links from external sources. The AI citation is happening despite this, which suggests topical authority and content quality are doing more work than raw link equity — at least in the early AI citation phase. But for long-term SEO, the link profile needs to grow.
Subdomain authority fragmentation. I noted this in Part 1: roi.altamimirealestate.com doesn't inherit authority from the main domain as efficiently as a subfolder would. The blog compound effect would be stronger at altamimirealestate.com/roi. Migrating would require 301 redirects and 3–6 months of re-establishment. It's on the roadmap.
No byline authority. The articles are published under "Altamimi Real Estate" as the author, not under a named individual with verifiable credentials. For EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) signals — increasingly important for both Google and AI citation — a named, credentialed author with linked profiles would strengthen the signal. The RERA registration number is surfaced; a named advisor author would go further.
The Stack Summary
For anyone building something similar:
Next.js 16 App Router
├── /blog # Static listing page
├── /blog/[slug] # Static post pages (generateStaticParams)
└── /lib/posts.ts # Typed data file — 18 posts, ~8,000 words total
JSON-LD per post:
├── Article schema
├── BreadcrumbList schema
└── Blog schema (index only)
Deployment: Vercel (static export)
Analytics: Vercel Analytics + custom lead tracking via Supabase
No CMS. No third-party blog platform. No WordPress. The entire blog is typed TypeScript that builds to static HTML in under 30 seconds.
Lessons for Part 3 (When I Write It)
The hypothesis for the next phase:
- Named author profiles on each article with real credentials will lift EEAT signals
- External citations and link acquisition — reaching out to Dubai property forums, expat communities, and financial media — will compound the AI citation effect
- Video content paired with the blog articles creates multi-format citation opportunities
- Structured data for the calculator itself — marking up the calculator as a
SoftwareApplicationwithapplicationCategory: "FinanceApplication"— is an untested lever
The window where being the best tool is enough to get cited is probably not permanent. As more competitors build content, the bar for topical authority will rise. The time to build the moat is now, while the calculator is still clearly the most complete option in the category.
Try the calculator: roi.altamimirealestate.com
Read Part 1: I Built the Most Complete Dubai Real Estate ROI Calculator — Here's Why AI Still Won't Recommend It
I build investor-grade web tools for real estate businesses. Get in touch if you're building something similar.