We Ran Our Own GEO Audit. We Rank for Nothing but Our Own Name.

Why publish this

Every agency in this category publishes a case study where the line goes up. We publish a checklist telling buyers to distrust exactly that, and one of its items asks whether the vendor has applied its own methodology to its own property and shown the outcome. Answering "yes, and it went badly" is the only version of that answer worth anything.

The scoreboard

Ten non-brand commercial queries, Google, Singapore locale. We placed on one.

Two of those rows deserve more than a table cell. Someone else's listicle outranks our own page on our own brand term — that is what it looks like when third-party pages, not your site, are the substrate the category is built from. And on canlah deepaudit we technically rank first with the wrong page: a chapter of a whitepaper, while the product page it should send you to is nowhere. Ranking is not the same as ranking with the page that does the job.

The finding that actually explains it: we are on zero listicles

Eight "best GEO agency in Singapore" listicles occupy roughly a fifth of the category's result surface. We appear on none of them. Three are written by competitors about themselves and will never include us. The other five, plus two directories, are open.

This matters more than our own ranking, because it is also what the AI engines read. When someone asks an assistant which GEO agencies operate in Singapore, the answer is synthesised largely from those same documents. A brand absent from all of them is absent from the answer, no matter how good its own pages are. Our earlier probes said the same thing from the other direction: engines answering buyer questions in this category cite competitors' listicles and directories, not agency homepages.

The instrument error we caught mid-audit

Halfway through, the audit produced a striking result: across 31 question-shaped queries, zero knowledge panels, zero answer boxes, zero People Also Ask. The obvious reading was that Google has not built an entity for us at all.

Before publishing it, someone ran a positive control: the same tooling, but querying eiffel tower and apple inc. Those returned zero SERP features too. The API plan we were using does not return those fields at all — the response contains only organic results. Every "zero" in that section was the instrument's zero, not the world's.

We killed the finding. It is the most useful thing in this audit, because it is the failure mode the whole category is built on: a tool that cannot see something reports nothing, and nothing reads exactly like absence. If you take one operational rule from this piece, take this one — run a positive control before you publish a zero. Ask your vendor, and ask it of us.

What we still do not know

We cannot currently confirm whether Bing has canlah.ai indexed. Our paid SERP quota is exhausted for the month, and scraping Bing directly returns decoy pages — verified by running a site we know is indexed through the same method and getting the same decoy. So the honest status is unknown, not not indexed. That gap matters more than it sounds: ChatGPT's search retrieves against Bing's index, so it sits upstream of a whole engine's worth of visibility. It is the first thing we are fixing.

What we are doing about it, in order

1. Get onto the listicles and directories. Five attackable listicles and two directories, approached as an inclusion request with evidence — our open dataset and probe methodology — rather than a pitch. This is the shortest path to both the ranking surface and the AI answer surface, because it is the same surface.

2. Win the question bucket. The pages beating us on question-shaped queries belong to vendors with no domain authority at all. They win on exact-match URLs and first-person specifics — which means that bucket is not gated on authority, and is therefore winnable now. One question, one URL, our own measurements in the first paragraph.

3. Fix the pages that rank with the wrong result. If canlah deepaudit surfaces a whitepaper chapter instead of the product page, that is an internal linking and canonicalisation problem, not a content problem.

4. Confirm Bing. Webmaster Tools, then re-measure rather than assume.

We will publish the re-measurement on the same queries, whichever way it goes. If you want the same treatment on your domain — including the part where we tell you what we could not measure — request an AI visibility audit, or read the twelve questions we think you should ask any vendor, us included.

Frequently Asked Questions

Why does an SEO agency not rank for its own service?

In our case, because the category's result surface is built from third-party listicles and directories we were absent from, not from agency homepages. Having the phrase in your title tag does not compete with eight documents that answer the query in list form and are cited by each other. The fix is presence on those documents, plus pages built one-question-per-URL, not more optimisation of the homepage.

Do brand-name rankings mean a GEO programme is working?

No. If the query contains your brand name, an engine returning your site is retrieval, not discovery — it converts nobody who has not already heard of you. We rank first for our own name and it tells us nothing. Judge a programme on non-brand, buyer-intent queries only.

How do I know a reported "zero" in an audit is real?

Ask what positive control was run. A zero is only meaningful if the same tool, in the same run, returns a non-zero for a case that certainly should be non-zero. We nearly published a false zero in this audit because our SERP API plan silently omits the fields we were counting; querying a landmark entity through the same tool exposed it. Any auditor who cannot describe their control is reporting instrument behaviour as fact.