Mapping local real-estate search demand with free autocomplete data: 9,647 suggestions for San Jose, Santa Clara and Sunnyvale

Yuki Nakazawa · October 2026

TL;DR: We pulled Google and Bing autocomplete for “city + real-estate topic” in San Jose, Santa Clara and Sunnyvale through the Egosaru API: 75 seeds per engine, 4,553 unique suggestions from Google and 5,094 from Bing. After cleaning, 60% of Google’s survived as local property terms, but only 10% of Bing’s. The kept terms show sellers vetting cash buyers, Santa Clara searches driven by “Santa Clara County” tax and assessment questions, and Bing users phrasing housing searches the way city programs name them (“city of Sunnyvale affordable housing”, “Sunnyvale BMR program”).

A local agent, broker or property manager who wants inquiries from the web has to answer a basic question first: what do people actually type? National keyword tools tend to hide the long tail of “city + something” phrases as “no volume”. So we ran the research step, collection to clustering, through the API of Egosaru, the autocomplete tool we operate. We first did this for three small towns in Hokkaido, Japan (English version, Japanese original). This is the same method run on a market English-speaking readers know, with the steps and the real numbers.

Why I care: I run an SEO agency, and I took over a real-estate company in Japan from its retiring owner. Keyword research for local property is part of my day job, not a side experiment.

Paid tools and free autocomplete are a combination, not a choice

I use Ahrefs every day. For search volume, keyword difficulty and the backlinks of ranking pages it is indispensable, and nothing here is an argument against paid tools.

But when a local business needs to enumerate thousands of “city + something” phrases, checking all of them in a paid tool gets expensive, and many are too small to be in its database at all. So split the work: cast a wide net with free autocomplete, then measure only the shortlist in the paid tool.

Paid tools also have usage limits. Once, after opening a large number of reports in Ahrefs in a short time, I got a “suspicious activity” notice and report access was restricted for a while. Paid tools are not built for brute-forcing thousands of phrases. Leave the brute force to autocomplete and spend the paid tool on the words worth measuring.

Autocomplete doesn’t tell you how often something is searched. It does draw the map of which word combinations people use. Draw the map first, then measure.

Step 1: city names alone return almost no real-estate terms

Google’s top suggestions for “san jose” are the Sharks, the time, the weather, San José State, the airport, the Earthquakes, and two other San Joses: Costa Rica and San Jose del Monte in the Philippines. “sunnyvale” gives the weather, the golf course, the farmers market, the library, and Sunnyvale, Texas. “santa clara” gives the university, the convention center and Santa Clara County. No property terms. Seed with city names only and most of the harvest is thrown away.

Google autocomplete for “san jose”: the Sharks, the weather, the airport, San Jose in Costa Rica and in Bulacan (Philippines). No property terms.

Bing autocomplete for “sunnyvale”: zip codes, the weather, the school district and Sunnyvale, Texas.

So the seeds were “city + real-estate axis”: 25 axes × 3 cities = 75 seeds.

Step 2: collect all 75 seeds through the API

For each keyword the Egosaru API returns its autocomplete plus the expansions with A–Z and 0–9 appended, so one seed is dozens of upstream queries handled for you. We collected Google and then the same 75 seeds on Bing, because more people reach Bing through Edge, Windows search and Copilot, and Bing words things differently.

“san jose homes for sale” on Google: 172 suggestions, including “by owner”, “with adu”, and the Zillow and Redfin searches we later filtered out.

curl -H "Authorization: Bearer $KEY" \
  "https://gstudio1.com/api/v1/google/suggest?q=san%20jose%20homes%20for%20sale&lang=en&flat=1"

A keyword nobody has fetched yet doesn’t come back immediately: you get HTTP 202, it joins the queue, and the same request sent again after retry_after_seconds returns 200 with the result. Results are cached for 7 days. The script only has to repeat “202: wait and resend; 200: save”. All 150 requests (75 seeds × 2 engines) were done in under two minutes. The spec is in the Suggest API documentation.

Depth varied a lot by seed and by engine:

Raw suggestion counts for the same seed on Google and Bing.

Bing’s big numbers for “sell house” and “land value” look like a gift. They aren’t, as the next step shows.

Step 3: throw away the noise

The raw lists can’t be turned into articles. Five kinds of noise dominated:

• Not about property. Expansions drift into jobs, sports and trivia: “santa clara real estate jobs”, “real estate agent san jose salary”, “san jose vs”. 976 Google terms; on Bing it was most of the list, 3,954 of 5,094.
• Another place with the same name. San Jose del Cabo, San Jose in Costa Rica, Batangas and Tarlac in the Philippines, San Jose in Illinois, Santa Clara in New Mexico, Sunnyvale in Texas (“land for sale sunnyvale tx”). 545 Google terms.

“san jose land value” on Google: Batangas and Tarlac, provinces in the Philippines, mixed into the suggestions.

• Portal and brand names. Zillow, Redfin, Craigslist, Opendoor and others. People searching “san jose homes for sale zillow” want Zillow, not your article. 135 Google terms.
• Fetch-location bias. Autocomplete depends on where the request comes from, not only on the language. Our collection servers are in Japan, and some suggestions came back with Japanese place names attached even with lang=en (“san jose home prices japan” on Google; Bing’s “san jose vacant houses in ikebukuro” is the clearest example). 165 Google terms, 68 Bing terms. If you run this kind of study, check for your own server location’s fingerprints.

“san jose vacant house” on Bing: Tokyo neighborhoods (Ikebukuro, Ueno, Akihabara) attached to a San Jose query. This is an artifact of where the request was sent from.

• City name dropped. The expansion lost the city (“how much do cash home buyers pay”, “space for lease near me”). 20 on Google, 217 on Bing.

The rules were simple: (1) keep terms with a property or transaction word (house, land, condo, rent, lease, vacant…); (2) drop anything with another region’s place name, a portal or brand name, or no city name; (3) a person reviews the final list, because rules miss things (our first pass let through ten terms with Japanese place names and a dozen job-search terms, which we then added to the filters).

After cleaning, 60% of Google suggestions survived and 10% of Bing suggestions.

Step 4: split by intent, then group into article units

We sorted what remained by what the searcher wants to do next: buy, rent, sell, check prices, find commercial space, choose an agent. Kept Google terms:

Kept Google suggestions by city and intent.

Autocomplete has no monthly volume, so the priority score is a stand-in: how many seeds a term appeared under, times its rank within the expansion. Terms with the same city, intent and leading words form a cluster, and one cluster becomes one article. “sunnyvale single story homes for sale 500k / 700k / 800k” is one article, not three; near-identical pages only compete with each other.

What the data said

Sellers vet the buyer. Google kept 543 selling terms. The “we buy houses” and “cash home buyers” seeds produced long tails about the buyers themselves: “san jose we buy houses scam”, “san jose we buy houses yelp”, “san jose we buy houses llc”, “san jose cash home buyers list”. The plain “sell my house fast” phrasing appeared only 3 times. An agent competing with cash buyers should write the comparison article (“cash offer vs listing in San Jose: what you actually net”) rather than another “sell your home fast” page.

“san jose we buy houses” on Google: “legit”, “llc reviews”, “sv ventures we buy houses san jose reviews”. Sellers are checking out the buyer.

Vacant property is a selling entry point here too. 111 kept terms contain “vacant”, for example “san jose vacant house permit”, “santa clara vacant house tax” and “san jose vacant house grant”. These are owners with a problem, which is where a selling inquiry starts. In the Hokkaido study the same pattern held: sellers searched for “vacant house”, not “sell”.

Santa Clara’s price questions are county questions. Santa Clara has 140 price terms, more than San Jose’s 135 despite fewer terms overall. 17 of them contain “county” and 23 are about tax or assessment (“santa clara property assessed value”, “santa clara county property value lookup”). Articles that explain the county assessor’s numbers, with a link to the official source, fit that demand better than a market-update post.

“santa clara home prices” on Google: county home prices, assessed value, property value assessment and property tax rate. Noise sits right next to them: “santa clara ivins” (Utah) and “osaka” (fetch-location bias).

Neighborhoods are searched by name. Downtown (31), South San Jose (15), West San Jose (12), North San Jose (9), Evergreen (8), Willow Glen and Japantown (6 each), Santana Row (4), Almaden, Cambrian and Rose Garden (3 each). That supports two levels: a city hub with neighborhood pages underneath, but only where there is something specific to say.

The employer map shows up. “santa clara apartments near nvidia”, “apartments for rent near nvidia santa clara”, “santa clara apartments near levi’s stadium”. Renters search by where they work. Rent rules show up too: “santa clara housing authority rent limits” and “santa clara housing authority rent increase”.

“santa clara apartments for rent” on Google: “under $1000”, “pet friendly”, and a run of Santa Clara Housing Authority terms.

Bing speaks the language of programs. Bing kept no selling terms at all; for this market Google is where seller demand shows. Google has some of this too (the housing-authority terms above), but the 371 Bing-only terms lean further toward official phrasing: “city of sunnyvale affordable housing”, “sunnyvale bmr program” (below-market-rate housing), “city of sunnyvale housing division”, “houses for rent santa clara county”. That is the opposite of Hokkaido, where Bing filled Google’s gaps in selling and commercial land. The lesson is the same either way: one engine is not the whole map, and which gap the second engine fills depends on the market.

Step 5: rules decided before writing

“City × something” can generate hundreds of pages mechanically. Thin pages that only swap the city name are what Google treats as doorway pages or scaled content, and they can drag down the whole site. So before writing anything:

• Every article carries at least one piece of first-party information about that area: county or city data with its source, the agent’s own transactions, photos taken on site.
• Every number has a source and a date. No home-made price estimates, no invented example listings.
• AI-written text is a draft only. A licensed agent reviews it before publishing, including fair-housing wording and the state’s advertising rules for licensees.
• Start with a pilot of about 10 articles, then watch Search Console and inquiries for 2–3 months before scaling.

Summary

Autocomplete research is more work in throwing away and grouping than in collecting. The flow has five steps: seeds from axes → bulk collection through the API → rule-based cleaning → intent classes → clusters as article units. In San Jose, 60% of Google’s suggestions survived and 10% of Bing’s, and the survivors changed the plan: “write for sellers comparing cash offers”, “explain the county numbers”, “build neighborhood pages only where people search them”, “use Bing for program and affordability terms”.

Paid tools measure; free autocomplete finds. Find widely and for free, and hand only the terms worth measuring to the paid tool.

Egosaru is built and run in Japan by Skyclear Co., Ltd.; the name is Japanese internet slang for searching your own name. The API covers 20 languages and Google, Bing, YouTube, Amazon (every marketplace), DuckDuckGo, Baidu, Yandex, Naver, Daum and more. The API key is free; see the API documentation and pricing. The web tool needs no sign-up: Egosaru.

更新履歴

• 第１稿投稿　2026年10月24日 1時14分（記事コンテンツアップ）