We Asked Gemini Why It Doesn't Recommend Us for Jakarta SEO.

We ran the query any prospective client would run. "SEO consultant Jakarta", straight into Gemini. It came back with a clean shortlist of six agencies and their specialties: Arfadia, cmlabs, ToffeeDev, Logique Digital Indonesia, Ematic Solutions, and Datamaya Consulting. Search Agency was not on it.

We run an agency whose entire thesis is winning visibility inside AI answers, and the most-used AI assistant did not name us for our own city and our own category. So we kept asking, and what came back is the clearest lesson in AI visibility we have seen in months.

The model knows us, in detail

The next prompt asked whether Gemini had any exposure to Search Agency. It did. The description was accurate enough to be flattering. It knew we were founded in 2019, knew the founder, placed us in Kembangan, Jakarta Barat, and framed us as an AI Search Optimization pioneer in Indonesia. It laid out our three layers of GEO, AEO, and core SEO. It named our AI Visibility Audits, the Search Agency Academy, the Weekend Course, and the SEO Fighter Bootcamp.

This was not a model that had never heard of us. It described our playbook more precisely than some inbound briefs do. The gap was never awareness. The model knew exactly who we are and still kept us out of the answer when someone asked for the category.

Why we were left off, in Gemini's own words

So we asked it directly. Three reasons came back, and none of them is wrong.

The first was boutique against enterprise. The shortlist had prioritized large, full-service agencies built to serve massive clients, and the model read us as a specialized boutique that takes on a narrower roster.

The second was the label. The phrase "SEO consultant" reads to the model as standard Google keyword and e-commerce optimization. Our work centers on LLM optimization, GEO and AEO, so the intent it inferred from the query did not match the box it files us in.

The third was our footprint. A large share of how we appear in Jakarta is educational, the Academy and the bootcamps, so the model partly sees us as the people who train in-house teams rather than a pure execution vendor.

Every one of those reads as fair. That is what makes it useful rather than annoying.

What this means for any brand

The lesson generalizes well beyond us. A recommendation is triggered by the intent a model reads into a query. Knowing your brand well does not, on its own, place you inside that intent. When someone types a category phrase, the assistant first decides what kind of answer they want, then fills it from the brands it associates with that intent. If your positioning has run ahead of the words the category still uses, you can be the best-known name for the new thing and absent from the query that still carries the search volume.

That is our exact situation. We pushed hard into AI search, the model learned us as an AI search and training brand, and the legacy phrase "SEO consultant Jakarta" routed straight past us to the agencies filed underneath it. One detail makes the point sharper: Gemini itself tags one of those six, Arfadia, as a GEO pioneer. The list is built on established category presence, not purely on who does the modern work.

Awareness and association are different problems

If you check a category prompt and your brand is missing, find out which problem you have before you spend a rupiah fixing it.

An awareness gap means the model cannot describe you, or gets your facts wrong. The fix is presence. Earn mentions and citations in credible places until the model has enough material to learn you properly.

An association gap, the one we have, means the model knows you well but files you under a different intent. The fix is reframing. Shape how the web talks about you so your name turns up in the category context you want to win, in the language buyers actually use for it.

Both look the same from outside, a brand that never appears in the answer. The work to correct each one is not the same, which is why guessing is expensive.

What we are doing with what we found

We did not trip over a flaw. We ran the test we run for clients, pointed it at ourselves, and it surfaced a real decision. Keep earning the modern query and let the legacy one route elsewhere, or do the work to be filed under both phrasings. We are choosing both.

The first step is the one we hand every client. Pick the category prompts you have to win, run them across ChatGPT, Gemini, Perplexity, and Google AI Overviews, and read whether you appear and which names sit beside you. Then close whichever gap the test exposes, awareness or association, with the work that actually moves it.

We could have left this unpublished. Running it in the open is the point. This is the job, and we found the gap on ourselves before we went looking for it on anyone else.

Our free AI Visibility Audit runs your brand and three competitors across the four major assistants on the prompts your buyers really use, and shows where you appear, where you do not, and which names the model offers instead. Talk to us and we will tell you whether the assistants are putting you on the list.