We ran our own answer engine optimization playbook on loudface.co, the same one we sell. Across the three AI engines we track, our share of the AI answers in our category went from 0.18% in April to 10.4% in June 2026, and we now hold the highest average cited position of any agency on our tracked leaderboard. This is the technical breakdown: the method, the measurement, and the parts that stayed hard.
For the short version with the narrative and the honest caveats up front, read the receipts. This page is the engineering log behind them.
The starting point: 0.18%
In April 2026 LoudFace was mentioned 8 times across 2,747 monitored AI answers in our category, surfacing in just 0.18% of them. That is the starting share of answer. For a buyer asking ChatGPT, Perplexity, or Google AI Overviews "who's the best B2B SaaS organic growth agency," we did not exist.
We had one structural problem that no amount of content volume fixes on its own: a Domain Rating around 31. In classic Google search, authority gates everything, and at DR 31 you wait years to outrank incumbents with a decade of links. So we did not bet on Google authority. We bet on the channel where page structure beats domain authority, which is AI answers, and we ran our own pipeline on our own site to prove the bet.
The method, step by step
None of this is secret. The work is the moat, not the method.
1. We ran the exact client cascade on ourselves. Every page went through the same loop we bill for: load the strategy and the live data, run SERP and AI-source reconnaissance, draft, run an anti-slop and voice pass, verify every claim against source, then ship. The cascade is the product, and running it on our own domain was the test.
2. We treated the top of every page as the citation surface. AI engines lift their answer from the first lines of a page rather than the body. So every cornerstone page opens with a direct answer of roughly 40 to 60 words that an engine can quote whole, uses question-shaped headings, and closes with a structured FAQ. We shipped FAQPage and Article schema on every post, because a page a machine can parse cleanly is a page it can quote.
3. We concentrated instead of narrowing. Our binding constraint was never which topics we cover. It was DR 31 gating the citation-to-ranking conversion, plus citation fragmentation, near-duplicate pages splitting the same answer across five URLs. The fix is consolidation: fold near-synonym pages into one canonical entity page so the authority and the citations pool in one place. We did that across the site rather than chasing more topics.
4. We built the pages AI actually cites. In our category the cited surface is decision content: "best organic growth agencies," "best AEO agency for fintech," agency pricing and comparison pages. We built that cluster deliberately, structured each page for extraction, and stamped every one with the year and a visible update date, because the engines reward freshness.
5. We picked targets from real demand rather than a keyword tool. Two signals drove the roadmap. The first was buyer prompts where every competitor scored zero, uncontested questions we could own outright. The second was our own AI-bot 404 logs: when ChatGPT's crawler fetches a URL on our domain that does not exist, the model has inferred a page should be there, which is a content brief that bypasses keyword research entirely.
6. We measured with a real stack and kept the metrics separate. We track share of answer in Peec across a fixed panel of buyer prompts, Google Search Console for clicks and impressions, and our own server logs, which are the highest-fidelity signal because they record what actually arrived rather than what a model estimates. We never conflate the three things most teams blur together: citation speed, visibility (how often you appear at all), and share of voice (how much of the answer is yours). Our headline number is visibility.
The result, broken out
The curve
Across April, May, and June, share of answer compounded.
| Month | Share of AI answers | Brand mentions | Avg position when cited |
|---|---|---|---|
| April 2026 | 0.18% | 8 | 3.3 |
| May 2026 | 3.56% | 330 | 3.6 |
| June 2026 | 10.44% | 1,245 | 2.9 |
The position number is the one we watch most. As visibility scaled, our average cited position improved from 3.3 to 2.9. The usual pattern is the opposite: as you fan out into weaker prompts, your average position degrades. Ours got better while volume grew, which means the engines are not just naming us more, they are naming us earlier in the answer.
By engine
Visibility is not evenly spread, and the spread is the actionable part (June 2026).
| Engine | Visibility | Avg position |
|---|---|---|
| Google AI Overviews | 14.21% | 2.9 |
| Perplexity | 11.36% | 2.7 |
| ChatGPT | 6.02% | 3.3 |
Google AI Overviews is our strongest surface, which fits how it works: it sits on Google's live index and updates within hours, so a well-structured page surfaces there first. Perplexity gives us the best position when it cites. ChatGPT is our weakest engine and our clearest growth gap, and it is also the engine that sends the most human traffic, so it is where the next quarter of work points. We wrote the engine-specific moves up as the ChatGPT playbook.
Which prompts we win
All of our tracked prompts are generic, non-branded buyer questions. There are no "is LoudFace good" prompts in the panel, so every number is competitive share earned cold rather than brand recall.
| Buyer prompt (June 2026) | Our visibility | Avg position |
|---|---|---|
| Best B2B SaaS organic growth agency 2026 | 60.3% | 1.5 |
| Top AEO agency for fintech 2026 | 55.9% | 3.1 |
| Webflow plus AEO agency for B2B SaaS | 47.1% | 1.8 |
| Best B2B SaaS marketing agencies for organic growth | 47.0% | 2.5 |
| Agency combining SEO, AEO, and Webflow | 45.6% | 2.2 |
On our core prompt, 60.3% of all answers now name LoudFace, usually first or second. We register on 46 of the 90 prompts we track, so 44 are still open. The runway is the prompts we have not covered yet, not a quality ceiling.
Which pages earned the citations
The citation engine is the decision-content cluster. The three most-cited pages alone pulled 745 inline citations in June.
| Page | Inline citations | Citations per retrieval |
|---|---|---|
| /blog/best-organic-growth-agencies-b2b-saas-2026 | 350 | 0.87 |
| /blog/best-aeo-agency-fintech-companies-2026 | 212 | 0.84 |
| /blog/best-b2b-saas-seo-agencies | 183 | 0.95 |
| /blog/webflow-agency-cost-b2b-saas-2026 | 169 | 2.64 |
The pricing and cost pages convert hardest: the Webflow cost page is cited 2.64 times per retrieval, quoted multiple times in a single answer when it surfaces. Pages that get retrieved but earn zero citations are the next structural fix, because they are being read and not quoted.
Where we sit against the field
We track 22 competitors. On raw visibility we rank 9th. On position we rank first.
| Agency | Visibility | Avg position |
|---|---|---|
| Omniscient | 23.18% | 3.2 |
| First Page Sage | 19.30% | 3.6 |
| Directive Consulting | 18.43% | 3.9 |
| Siege Media | 16.12% | 4.2 |
| LoudFace | 10.44% | 2.9 |
At an average cited position of 2.9, we are named earlier in the answer than every larger competitor on the board, including agencies appearing in twice as many answers. Closing the visibility gap is a coverage problem, more prompts answered well, rather than a quality problem, because the placement is already best in class.
The honest part: Google stayed small
This growth happened in AI answers rather than Google clicks, and that was the plan. Over the same 180 days, loudface.co drew 847 organic Google clicks at an average position of 26.8, with impressions roughly half what they were in December. On Google we are still on page three.
That gap is the whole thesis. Raw Domain Rating does not decide who gets cited in AI answers; structure, freshness, and specificity do. So a DR-31 domain can take 10% of its category's AI answers while its Google footprint stays capped, because authority still gates the citation-to-ranking conversion in classic search. AI search is where a smaller brand can win now, by being built to be quoted. Google compounds slowly behind it. For how the timelines differ, we broke down the three speeds of AI citations separately.
Generic AEO advice versus what we shipped
| Common advice | What we actually did |
|---|---|
| Publish more content | Consolidated near-duplicate pages into canonical entity pages |
| Add schema everywhere | Front-loaded a 40 to 60 word answer at the top of every page, then schema |
| Chase high-volume keywords | Targeted buyer prompts competitors scored zero on |
| Optimize for ChatGPT first | Optimized for Google AI Overviews first, the fastest surface |
| Track rankings | Tracked share of answer, position, and AI-referred sessions as separate metrics |
What this means for your program
The transferable lesson is that AI-answer share is winnable before domain authority is, if the pages are built to be quoted and the work continues through the quiet first quarter. The first stretch of any AEO program is invisible by design: the foundation produces nothing screenshot-worthy for weeks, which is exactly why most teams quit before the curve turns up. We wrote about the invisible first quarter because we lived it on our own site.
If you want to see where you stand today, run a free AI search visibility audit. It shows your share of answer across ChatGPT, Perplexity, and Google AI Overviews, the same baseline we started from at 0.18%. If you want a team to run the program, here is how we run AEO and what it costs, starting at $5,000 a month.
Limitations, stated as method
We would discount this if a competitor published it without these, so here they are. The sample is one brand, our own, over three months, and June is a partial month, so the percentages climb off a near-zero April base. Our panel covers three engines, ChatGPT, Perplexity, and Google AI Overviews, rather than every AI surface. Visibility is the share of answers that mention us; share of voice, the share of the answer that is ours, is a stricter and lower number. The prompt panel grew over the window, so month-over-month visibility is measured against a moving base. And Peec's citation counts are a modeled daily sample rather than server logs, which is why we cross-check against our own logs. None of this changes the direction. It changes how loudly we can claim the size.
