A Los Angeles apartment broker, eight months of nothing, and a batch that thinned out

If you read nothing else

• For eight months from September 2025, the site earned 12 to 57 clicks a month from Google.
• A large build landed in May 2026. Monthly clicks went to 110, 137, 229, then 283 in August, and then eased to 234 in September.
• In the latest two 28-day windows, pages at position 10 or better rose from 712 to 813.
• A batch of 238 pages published in August showed up fast and then thinned out. 197 of them appeared in the peak week. By early October only 72 appeared. The ones that kept appearing held position seven. They did not slide. Most simply stopped appearing.
• On an AI answer check, ChatGPT named the firm on four of eight questions. Claude named it on one and linked to it on four.
• This page does not give a deal count. The firm's own published figures are not consistent across the places it lists them, and the principal has not ruled on which is right.

1. The result, in sixty words

Between September 2025 and October 10, 2026, stermanmfg.com went from about 40 Google clicks a month to a peak of 283 in August, then 234 in September. Pages at position ten or better rose from 712 to 813 between the latest two 28-day windows. The site is more visible and is visited less. This page says why that is not a contradiction.

2. Who this is

Sterman Multifamily Group is a Los Angeles brokerage that sells apartment buildings, run by a single principal who has worked in the market for decades. It is a corporate-affiliated broker with a personal brand: a seller choosing someone is choosing a person.

The goal was to be found by building owners who are weighing whether to sell, and by the people who advise them: asking who to hire, what the rent rules mean, and what a sale would cost in tax.

The question this page answers: can a single-broker practice in a high-trust, high-value local market earn a position in search for the questions owners ask before they ever pick a broker, and what does that look like month by month?

3. What we were up against

A decision that is made on trust, rarely on a search

Few people hire an apartment broker because a search result looked good. The searches that matter are early questions (rent caps, transfer tax, what a building is worth), and the people who ask them are not yet buyers of anything.

A market full of established names

Los Angeles is crowded with national brokerages and well-known local teams, many with large review counts that show up beside their results. ### The firm's own record, stated in several places The firm publishes its transaction record in more than one place, and the figures do not match. We chose to keep a total off this page entirely until the principal rules on which figure is right. A record that is quoted three ways is not a record a case study should quote a fourth time.

A young site

The site began its measured life in September 2025 with a handful of pages. A large part of what we measure here is a new site building a history.

What it had going for it

A real practice with a real track record, a principal willing to answer hard questions, and a market where the questions owners ask are specific, regulated and underserved.

4. What was shipped

Counts by kind and date only.

• A large build in late spring 2026 (hundreds of pages in one release), then a steady run of additions through the summer.
• August 2026: 238 pages, the batch this page looks at most closely.
• By October the site held about 1,400 pages, including neighbourhood market pages, closing records, seller guides, glossary entries and a map.
• Pages were built from the firm's own transaction records and public sources, and carry their sources.

We are not describing how pages were chosen, built or checked, or in what order.

5. Month by month

The first eight months set the baseline at 12 to 57 clicks, and almost all of it was people typing the principal's name. From May, most visible searches were not the name. A wider audience clicks at a lower rate, so the click-through rate fell from 2.5% to under 1% while clicks rose. A falling rate in this case is not a decline.

Latest two equal 28-day windows

The site is easier to find and is clicked less. The next two sections are about why.

6. The August batch

Of the 238 pages published in August, 236 appeared in Google at some point.

What happened: 197 of these pages appeared in the peak week. 63 appeared in the week of October 5. 134 did not.

What did not happen: the pages did not slide down the results. Among those still appearing, the median position was 6.8 at the peak and 7.0 now. Only 13 of the 63 are more than three places worse. Fourteen are better.

What it looks like instead: most of the 197 appeared in only two to four of the nine weeks. Eighteen appeared once. Pages that are shown only occasionally are being tested, and some are then shown less. A very rare search can only bring a page into view when someone types it.

By page type, of those shown at the peak and shown again on October 5:

• the neighbourhood market page: 46 of 133
• the neighbourhood guide ("living"): 9 of 17
• the seller guide ("selling"): 3 of 17
• the FAQ: 4 of 16
• the closings record: 1 of 14

The main market pages held. The supporting pages thinned out most.

7. AI answers

Four assistants were asked eight questions, locked before the first run, three samples each. Four were about choosing a broker and four were about the rules and costs of selling.

Set A, choosing a broker: best broker to sell an apartment building in Los Angeles · multifamily broker Los Angeles sell side · who should I hire to sell my rent controlled apartment building in Los Angeles · how to choose a multifamily broker to sell my apartment building in Los Angeles

Set B, the rules: what is the RSTPO rent increase cap in Los Angeles County · what is the AB 1482 rent cap in 2026 · how does Measure ULA transfer tax apply when selling an apartment building in Los Angeles · how do I sell a rent controlled apartment building in Los Angeles

Result, October 13:

• ChatGPT named the firm on 4 of 8 questions and linked to it on 4. It named it on three of the four broker questions.
• Claude named the firm on 1 of 8 and linked to its pages on 4 of 8, including all three samples of the rent-cap and sale-process questions.
• Gemini and Perplexity returned errors on every question, so they are missing, not "no".
• Not found by either assistant: "how to choose a multifamily broker", and the AB 1482 question.

One pattern: the assistants cited the firm's pages as sources for the rules (Claude) and as a broker to consider (ChatGPT). They did not agree about which role it plays.

Limits: eight questions we wrote, one day, three samples. A link is not a recommendation.

The long-tail check, October 13

The eight questions above are broad. People who need a rent-control sale ask narrower ones, so we ran a second set built from the firm's real Search Console searches and the rules it writes about.

Where the firm is cited (one sample each, October 13. ChatGPT logged out, Claude through its API, Google AI Mode in the owner's signed-in browser, which may favor sites it has visited):

The pattern: the firm is picked for the specific job, a rent-controlled building, and the rules around it, not for "multifamily broker". ChatGPT names it as a broker. Claude links its pages as the source for the rules.

How these were chosen: from the firm's real searches and the topics it publishes on. The table shows the questions where it is cited, so read it as a list of where it appears, not as a random sample of everything a seller might ask.

For how assistants choose which businesses to name, see how AI assistants pick businesses and what generative engine optimization is.

8. What went wrong, and what did not move

• The August batch lost two thirds of its visibility in five weeks, and clicks to it fell from 16 a week to 5.
• Publishing stalled twice: about a day and a half around September 1, and about four days from September 5 to 9, because of faults in how pages were released. Both were found, fixed and guarded against. The stall of September 5 to 9 sits inside the weeks the August batch was thinning.
• The live site stopped updating for about a day on October 12 after a change in where the code is stored. It was fixed on October 13.
• A page we published carried a stale total in its footer and its structured data after the record had changed. It was found in a sweep and corrected.
• The firm's public review count is a gap, and content cannot close it.
• No measure of enquiries exists. Clicks are not mandates.

9. Why it behaved this way

Why a batch can appear fast and thin out. A search engine tries new pages in front of a few searchers to see how they do. For a rarely typed question, a page can only appear when that exact question is asked. Pages answering narrow questions may be seen a few times and then not again for weeks.

Why position held while appearances fell. Position describes where a page sat when it was shown. If it is shown less often, the number does not change, because there are fewer showings, not worse ones.

Why the supporting pages faded most. A closings record or an FAQ answers a narrower question than a market overview, and fewer people ask it.

Why clicks can fall while top-10 pages rise. More pages near the top is a statement about presence. Clicks depend on how often someone types a search that a page answers. A busy period for typing a name, such as late summer, can lift clicks for reasons that have nothing to do with pages.

What best explains what we saw, ranked by how much each would change the conclusion if proven:

1. How seasonal this business's search is. The traffic is weekday, business-hours and desktop, and we have no previous year to compare.
2. How much of the August batch was a trial period that a search engine gave and then withdrew.
3. Whether the supporting pages are worth keeping or are better folded into the market pages.

Terms used here are defined in the glossary, and the broader method is in the Search Playbook.

10. What we believe, and what this case did to it

This section is opinion.

Be the answer, not the ad. The pages that held were the ones that answered a market's main question completely.

It still takes time. It shouldn't take faith. We are publishing the batch that thinned out because a result you cannot audit is not a result.

Search did not shrink. It spread out. A firm that was barely found a year ago now appears for more than a thousand pages' worth of searches and is clicked less than it was in August.

What we refuse to do: quote a deal total the client has not agreed, count a rise in appearances as a win, or call a month a trend.

What this case changed our mind about: we expected more visibility to mean more visits. Here it did not, for weeks.

11. What we are not claiming

• That any of this produced a mandate or a sale. Search Console does not measure enquiries.
• That the August batch's thinning will continue, or that it will reverse.
• That the AI-answer result applies to Gemini or Perplexity.
• That the firm's transaction record is any particular number.
• That the same would happen for another broker.

12. What we are watching next

Dated, and written before the result.

• 28 days to November 8: clicks of at least 200, or we say they did not.
• By November 10: at least 100 of the August batch's pages appearing in a week, or we say they did not.
• At the same time: the same eight questions on every assistant that is working.
• When the principal rules on the deal total: the record page and this page are updated together.

13. How to check these numbers

Search Console, property stermanmfg.com (domain), Performance → Search results, Search type Web, date range September 25, 2025 to October 10, 2026, and the weekly page counts come from the Pages tab per week.