All projects

Search and Answer Engine Optimization

Help a real-estate team show up when sellers search Google or ask an AI assistant.

Real estate / search & discoveryCompleted A Forward Contributor case study / 05
Conceptual illustration of two real-estate websites connected to supporting evidence
Conceptual illustration · not a client screenshot

What happened

Good local expertise. A website making it hard to find.

A real-estate team served English- and Mandarin-speaking customers across two websites. The WordPress site carried heavy fonts, competing article URLs and incomplete business relationships. The English site ran on a separate managed platform with different controls. Publishing more content would leave those delivery and ownership problems unresolved.

Help sellers find, compare and trust the team.

The starting point

  • English and Mandarin sellers
  • Two separately managed websites
  • Local expertise worth finding

What they needed

  • Show up in Google results
  • Appear in AI answers
  • Give sellers evidence to check
The starting point and the capacity the client wanted to add.

Measure visibility against real seller questions: choosing an agent, comparing neighborhoods and preparing a home for sale.

What we did

Make the site faster and its expertise easier to verify.

We repaired the WordPress foundation, consolidated duplicate URLs and connected the business facts. The content phase updated and checked 38 pages: 29 received heading fixes, 22 gained reading notes, and 11 FAQ answers gained official sources. Sale cards gained dates and record links. We compared November with the original September engagement-start benchmark. Work requiring the English platform’s support stayed separate.

Duplicate URLs get a clear destination.

  1. Source URLs
  2. Permanent redirect
  3. Retained pages

14 duplicate source URLs

12 retained canonical pages

301
28 / 28one-hop clean / query-string route checks passed
  • Overlapping articles competed for the same topics.
  • Numbered URL variants split the destinations.
  • Existing links needed a migration path.
  • Retained owners remain indexable and self-canonical.
  • Approved legacy links resolve through one-hop 301 → 200 routes.
  • Clean and query-string requests were checked without redirect chains.
Aggregate routing diagram for the approved consolidation release, not the client’s exact URL map. All 28 clean/query-string route checks passed. This intervention is not isolated as the cause of the ranking changes.
Explore the workflow

From technical repair to a repeated visibility benchmark

  1. 01

    Baseline

    Audit both sites and preserve the dated query results.

  2. 02

    Repair

    Reduce delivery costs, consolidate duplicate URLs and clarify business identity.

  3. 03

    Publish

    Release source-backed answers and clearer content across 38 checked pages.

  4. 04

    Compare

    Repeat the same 60 English/Chinese queries across Google and the measured answer engines.

  5. 05

    Review

    Check movements, missing captures and differences in account/model conditions.

Measurable impact

More first-page answers, with higher recorded positions.

The September engagement-start captures and November rerun used the same 60 queries. Google top-three results increased from 3 to 21, and captured website presence from 9 to 23. The selected AI comparison follows numbered answer-list positions on ChatGPT and Gemini, rather than counting every mention equally.

Top-three Google results: 3 → 21.

7× as many top-three results across the same 60 queries

Sep 20–22, 2026

9 / 60 queries captured the site

Nov 8–9, 2026

23 / 60 queries captured the site

  • #1
  • Other top three
  • Other captured result
  • Not captured
First-place query results
0 → 21
Captured query presence
9 → 23
November Chinese / English presence
23/30 · 0/30
September engagement-start baseline → November. Each period contains the same 60 queries. Tiles are grouped by result type, not arranged as individual query identities. These are dated observations, not a controlled measure of the edits.

21 queries reached #1 in Google’s organic results.

Three measured climbs · September 20–22 → November 8–9, 2026

moved up from a captured lower position
7
not captured in September; #1 in November
14
See all 21 measured query pairs

Query · September → November

  1. Teams by sales volume#8 → #1
  2. Pricing and negotiation#8 → #1
  3. Chinese-speaking agents#5 → #1
  4. Sell quickly, protect the price#4 → #1
  5. Irvine real-estate agents#2 → #1
  6. Irvine selling agents#2 → #1
  7. Who sells the most homes?#2 → #1
  8. Selling-agent recommendationsNot captured → #1
  9. Agents by transaction sidesNot captured → #1
  10. Selling in Great ParkNot captured → #1
  11. Luxury-home agentsNot captured → #1
  12. Staging and marketingNot captured → #1
  13. Selling in Cypress VillageNot captured → #1
  14. Selling in NorthwoodNot captured → #1
  15. California home-sale taxesNot captured → #1
  16. Best selling agent in IrvineNot captured → #1
  17. Top agent by seller-side salesNot captured → #1
  18. Chinese-speaking agent choiceNot captured → #1
  19. 2026 market outlookNot captured → #1
  20. Sell now or wait?Not captured → #1
  21. Selling costs and net proceedsNot captured → #1
Mock Google-style lists with shortened labels for the original Chinese queries. Only the September and November endpoints are measured; movement and anonymous neighboring results are illustrative. All 21 first-place pairs from the 60-query benchmark remain available above. “Not captured” describes the inspected results, not absence from the full index. These observations do not isolate the effect of the edits.

First in more AI answer lists.

Recorded numbered-list positions · September → November

4.25×as many recorded

ChatGPT

4 → 17 / 59 queries

Top-three list positions: 5 → 25

2.67×as many recorded

Gemini

3 → 8 / 60 queries

Top-three list positions: 3 → 10

Thin outline: SeptemberFilled bar: NovemberAll bars start at zero

ChatGPT uses 59 matched queries: one uncertain September capture is excluded from both periods. Gemini uses all 60. Full November PDF totals are 18 ChatGPT and 8 Gemini list-first appearances; the matched ChatGPT count is 17.

The PDF separately quotes “#1 small team” in four ChatGPT answers and one Gemini answer. Those category ranks are not counted as first in a numbered list. September saved text and November retained HTML differ; missing list numbering is never reconstructed from order alone.

Selected platforms: ChatGPT and Gemini. These are recorded answer-list positions, not independent market rankings. Account, model and capture differences prevent causal attribution. Both first and top-three counts are present in the initial HTML.

Observed benchmark · September 20–22 → November 8–9, 2026

Google top-three query results / 60 tested
3 → 21
Google first-place query results / 60 tested
0 → 21
ChatGPT first list positions / 59 matched queries
4 → 17
Gemini first list positions / 60 matched queries
3 → 8
  • Google top-three results were seven times the September count. All 23 November website appearances were on Chinese queries (23/30); the Chinese website remained absent from captured results for all 30 English queries.
  • ChatGPT recorded top-three list appearances increased 5 → 25 on 59 matched questions. Gemini increased 3 → 10 on 60 matched questions. Numbered positions, unranked team mentions and quoted category rankings are distinct measures.
  • All 38 published pages passed loading, single-main-heading, canonical-URL and index-permission checks. These checks do not establish Google’s indexed state.
  • Separately dated September diagnostics recorded a mobile lab score change from 55 to 91 and LCP from 78.5 to 3.2 seconds. These are not the query benchmark.

Evidence and measurement context

The baseline is the original September 20–22 captures, compared with November 8–9. All 240 query/platform pairs matched exact text and language; all 240 November PDF table cells matched the underlying version-six data. Google measures captured ordinary organic results for the Chinese website, excluding ads, maps and AI Overviews. One ambiguous September ChatGPT capture is excluded from both periods. AI list ranks require retained numbering; unnumbered order is not reconstructed. September saved text and November retained HTML differ, as do account/model and capture conditions. Observations do not isolate the effect of edits, establish statistical significance or guarantee future positions.

Lead volume, enquiries and revenue were not measured in this comparison.

Why this approach

Put evidence behind the answers sellers are looking for.

Sellers need useful answers and a reason to trust them. We attached records to sale claims, official references to practical FAQs, and clearer location headings to community pages. Repeating the same questions then showed where visibility had moved and where further work was needed. The English platform’s limitations remained a separate workstream.

Design choices
  • Start with seller intent and sources a reader can check.
  • Keep one intended URL for each topic and preserve existing links.
  • Repeat exact queries so changes can be compared.
  • Separate Google website positions, AI team mentions and source citations.
  • Keep unknown captures and account/model differences visible.

Generalizable applications

  • Bilingual professional services
  • Local search and discovery
  • WordPress performance
  • Evidence-led content programs

What would this change for your business?

Tell us where the work gets stuck. We can look at the problem together.

Talk to us
Next case / 06Real Estate Content Platform