We'll be upfront about something before we get into the numbers.
Client work came first this month. It always does. We're a small (yet agile and experienced) team running a live experiment on ourselves while delivering real work for real clients every day. Some weeks the experiment got squeezed into the gaps. If this post is a week later than we said it would be - that's why.
But here's what we learned from that: even when we weren't actively working on the experiment, the work we'd already done kept on working. We'll take that.
When we started this experiment in May 2026, our overall brand visibility in AI search was 0.21%. We were ranked for 8 of 50 target keywords, all brand-name terms. Our share of voice against competitors was 1%. Sad faces all round.
Here's where we are now.
Overall brand visibility: 8.51% - up from 0.21% at baseline. That's a 40x increase in six weeks with zero paid media. If you're curious about your own brand, you can check your visibility for free using HubSpot's AEO Grader.
ChatGPT visibility: peaked at 21.7% on 22 June, then pulled back to 13.7% by 29 June. We'll talk about that pullback - it's one of the most interesting things we observed.
Gemini visibility: 2.3%, up from zero. Still modest but consistently moving in the right direction.
Perplexity visibility: 2.2%, up from zero. Still growing week on week at the time of publishing this blog.
Share of voice: 12%, up from 1%. A major competitor dropped from 58% to 45%. We're now the 6th most cited domain in our competitive category - sitting above several agencies with far more established online presences.
We've done weird happy dances you probably wouldn't want to see.
Here are the five things that made the measurable difference - in order of impact based on what the data shows.
1. Organisation schema on the homepage
This was the most urgent fix and it worked faster than anything else. Before we added it, AI engines were genuinely confused about who BLOOM Transitions is - regularly surfacing a US youth nonprofit with a similar name. Within days of the schema going live, that confusion was resolved. A disambiguation fix is the single most important thing any business can do if AI is getting their identity wrong.
2. A dedicated AEO services page
Publishing a page specifically structured for AI extraction - with direct answer copy, FAQPage schema, and eight Q&A pairs - gave AI engines a clean, citable source for AEO-related queries in the NZ market. This page is now being cited directly. It also answers two of our eight tracked AI prompts simultaneously.
3. HubSpot Marketplace listing rewrite
HubSpot.com is the most cited domain in our competitive category by a significant margin - 2,083 citations in our tracking data versus 468 for the next highest. Getting our listing copy right so AI engines can extract accurate information about BLOOM from that authoritative source matters enormously. The rewrite was straightforward but the impact has been real.
4. A Vanguard 86 Peer Relationship
We reached out to Vanguard 86 - one of the most active NZ HubSpot agencies and a significant source in our citation data (at an impressive 466 citations). They responded warmly, agreed to include BLOOM in their NZ HubSpot agency roundup, and proposed a reciprocal relationship for client referrals outside each other's ICP. Vanguard 86 currently sits as the 3rd most cited domain in our category. Being mentioned there will be a meaningful citation source when it goes live. We are super grateful for our HubSpot community.
5. English (NZ) language settings
A small technical change - updating our site language attribute to en-NZ - sends a geographic relevance signal to AI engines that's consistent with our schema's areaServed field. It's one signal among many, but consistency across signals is what builds AI confidence.
Our ChatGPT visibility peaked at 21.7% on 22 June then dropped to 13.7% the following week. We didn't change anything between those two dates.
This is normal AI behaviour and worth understanding. AI models don't update continuously - they have training cycles and weighting adjustments that can cause visibility to fluctuate even when your content hasn't changed. The pullback doesn't mean the work stopped working. It means AI visibility isn't a straight line upward, and anyone who tells you otherwise isn't being straight with you.
The more important signal is the trend over six weeks - which is consistently and significantly upward across all three platforms.
This is the third data source we're tracking alongside the HubSpot AI grader and citation analysis - and it tells an equally encouraging story.
Our SEMrush position tracking is filtered here to keywords ranking in positions 1–3. In May, BLOOM ranked for 8 keywords total, all brand-name terms, none with any meaningful commercial intent. Here's what's changed by 4 July:
"hubspot audit nz" - position 1, AI Overview
This keyword had a dash in our May baseline - meaning BLOOM was not ranking at all. We're now ranking number one and triggering a Google AI Overview, meaning BLOOM is being cited by Google's own AI for a commercial HubSpot query. That's the AEO and SEO work crossing over into traditional search in exactly the way we hoped.
"what is bloom transitions and what services do they offer" - position 1, AI Overview
This is the disambiguation prompt we've been tracking since day one. In May, AI engines were confused about who we are. That prompt now returns BLOOM at position 1 with an AI Overview. The disambiguation fix worked completely.
"bloom transitions digital agency" - position 1, AI Overview
New ranking with a visibility increase of 7.3%. This term didn't exist in our tracked set.
Three tools - HubSpot AI grader, citation analysis, and SEMrush - all pointing in the same direction. The foundation work is being recognised across AI search, traditional search, and citation tracking simultaneously. If you don't have the access or bandwidth to measure these yourself, you need someone that does.
The foundation phase moved faster than we expected. The authority building phase is going to take longer - and we're being honest about that.
Client case studies - we haven't published a single structured case study yet. Case studies are high-trust citation sources for AI engines. One well-built case study with direct result statements and FAQPage schema will outperform ten blog posts for citation purposes. This is our next priority.
Peer citation outreach - we've identified the sources we need to be on. We haven't systematically pursued them yet. Reddit and YouTube are both appearing in our citation data as UGC sources - channels we haven't touched at all. Yet. That's a gap worth addressing in Phase 3.
Client reviews on HubSpot Marketplace - in progress. Reviews from verified HubSpot clients directly influence how AI answers "best HubSpot partner NZ" type prompts. We've sent the requests. We're waiting.
Commercial keyword rankings - while the top 1-3 data is encouraging, we haven't yet seen movement on the harder commercial terms - "hubspot partner nz," "aeo agency new zealand," and "seo agency new zealand." These are the terms with real buyer intent and meaningful search volume. They take longer. Post 4 will include the full keyword ranking picture across all 50 tracked terms.
If you take nothing else from this post, take this: the foundation work moves faster than most people expect. Schema markup, entity clarity, and structured content are not six-month projects. They're weeks - and the results can be visible almost immediately.
The harder, slower work is building the external citation profile that tells AI engines your brand is trusted by sources they already trust. That's what we're focused on now.
If you're curious whether your brand has the same gaps we started with, the fastest way to find out is to ask ChatGPT or Gemini: "What is [your business name] and what do they do?" Then ask: "Who are the best [your service] providers in [your region]?"
If you don't like what comes back - or you don't appear at all - we can help.
Keep following the experiment on LinkedIn - post 4 will land in four weeks with month-two data and a more detailed look at what the citation data is telling us.
To be honest on the timing; we’ll try our best, but clients will always come first.