From 0% to 18% AI visibility in a brand new market.

Humand builds HR software for deskless and frontline workers. AI assistants already recommended it in Argentina and had never once named it in the United States. The same weekly system moved both.

Visibility by market, every week

Share of tracked AI answers naming Humand, 15 April to 3 September 2026

1 2 3 4

Timeline

• Apr 15 1 English roundup published
• May 13 2 Argentina tracking begins
• Jun 10 3 Reddit comments live
• Jul 8 4 Spanish guides live

24%→24%

Argentina AI visibility across 16 weekly reports, measured on six AI engines with prompts that never change. Report by report figures are in the chart above.

0% → 18%

US AI visibility. First report in that market scored zero.

4% → 11.3%

Argentina citation share, Across 8,106 citations recorded in the window.

A category leader at home, absent from the answers abroad

Humand builds a mobile-first employee experience platform for deskless and frontline workers: the people in warehouses, on shop floors and out in the field who have never had a work laptop. Over 1.5 million of them use it, across 1,500 organisations.

In February 2026, the company raised $66M, with US expansion named as one of the things it was for. Ask an AI assistant in Spanish for an employee experience platform for plant and field staff and Humand came back in about a quarter of the answers. Ask the same question in English, for the US market, and it came back in none of them.

That is a harder problem than a low score. A quarter of the answers is a position to defend. Zero is a category that has already been decided by other people, in public, before the company arrived.

Their rivals were already writing the category

Before writing anything, every source cited in a tracked answer was pulled and ranked. That was 13,260 URLs across 3,598 domains. Eight of the twelve most cited were rival vendors publishing about their own market, each carrying five hundred citations or more. humand.co sat forty-sixth with 174.

What the answers were actually citing

The number that set the plan:

YouTube was the most cited domain in the category, and those 1,145 citations were spread across 532 different videos. The single most cited one was reached 56 times.

The work

Four moves, in order

Every one of them came out of the citation data. None was a guess about what might work.

01 Defined the market before defining the content

Two markets, two prompt sets, two reports, running side by side every week. The prompts were written in the language buyers use in each one, and then they stopped changing, so every week after that is a comparison rather than a fresh measurement.

02 Published the thing the answers were already reaching for

Eleven pieces are live across four markets. Two of them carry most of the load.

03 Commented in the Reddit threads AI actually cites

Reddit was the third most cited domain in the market. Every thread cited anywhere in the account was ranked by citation count, and the top of that list got the work: eight comments, across seven subreddits.

04 Let the weekly report pick the next piece

Nothing here ran on a content calendar. Each week's report named the prompts that had moved and the prompts that had not, and the next piece was written against whichever gap was widest.

The outcome

The new market opened, and the old one doubled

Where the movement came from

Change in visibility per engine, over each market's tracked window

• Perplexity Sonar +32.5%
• Google AI Mode +25.8%
• ChatGPT Search +10.8%

At home, their own site is now the second most cited source in the category

Across 8,106 citations recorded in the Argentina window, humand.co is reached more often than any competitor domain and is beaten only by YouTube. Citation share went from 4% in May to 11.3% at the end of August.

And in the US, the answers name them

Two Google AI Mode answers, three weeks apart, on two different buyer prompts. Both rank Humand inside a shortlist of platforms that were already there in April, when the company was in none of them.

What's next

The written half is working. The rest is still open.

1. Take the video slot.
2. Earn the outside mentions.
3. Close the schema gap.
4. Run the pattern into Mexico and Colombia.

Why this worked. Tracking each market separately, every week, on prompts that never change is what made the zero visible at all. One blended score would have averaged it away.