TL;DR
Gladia's organic performance was worsening as developers moved to AI search, and they were unsure whether content written for a technical audience would hold up. Discovered Labs put a structured onboarding in place to get content live in the first couple of weeks: technical, SEO and AI visibility audits, a 90-day pillar roadmap they could see 30 days out and developer-grade content run end-to-end in their CMS. Over four months, monthly sales-accepted leads grew sevenfold and lead quality rose alongside volume, with AI search emerging as the largest new driver on top of a stronger SEO foundation.
Hear directly from Anna on how the work changed Gladia's pipeline and AI search visibility.
Organic performance was worsening, and nobody could tell which AI search tactic actually worked
Gladia builds AI audio infrastructure, the audio APIs developers wire into their products for transcription and audio analysis. For most of the company's growth, organic search carried inbound. It was, in their words, the bread and butter.
Then the channel started to move. Website traffic was declining, and the way developers find and compare audio APIs was shifting toward AI assistants. They were asking ChatGPT which API to use instead of scrolling ten blue links. Organic, Anna told us, was really not what it used to be.
The harder part was the noise. AI search was new enough that everyone had a theory and no one had proof. Every week brought a fresh way to get cited: restructure for the LLM, add this schema, get mentioned on Reddit, chase a different metric. None of it came with evidence that it moved anything. For a team deciding how much to publish and what to fix on the technical side, there was no reliable way to tell real advice from someone guessing out loud. And Gladia sells to developers, a naturally skeptical audience, so whoever they brought in had to write for engineers and work alongside their content team.
“We felt like we could really bring a specialist on board who would help us isolate the signal from the noise, and adopt the best strategy in order to adapt to this paradigm shift.”
Anna, Head of MarketingThe system: an end-to-end AEO and SEO engine, live in weeks
At Discovered Labs we work exclusively with B2B SaaS companies, so we understand the pressure put on marketers. Our client onboarding is designed to capture commercial opportunity as fast as possible rather than getting bogged down in weeks of audits and planning. Here's what that looked like for Gladia.
1. Audits and diagnosis: finding gaps and opportunities
We complete a mixture of audits for clients across both SEO and AEO surfaces. For Gladia, this consisted of a technical audit, an on-page audit and an AEO audit using our in-house AI visibility tracking platform. Our technical audit surfaced indexation, performance and crawlability issues, most of which were handled by our in-house engineers.
On the AI search side, we mapped where Gladia showed up versus competitors across ChatGPT, Claude, Perplexity and AI Overviews. For the first time, Gladia could see which buyer questions they were invisible for, measured against a fixed prompt set rather than a score that drifts week to week.
You can learn more about our AEO measurement methodology on our website.
“One big part of this that we didn't necessarily think about would be as substantial as it was. That really made a difference.”
Anna, Head of Marketing2. The 90-day roadmap: a diagnostic-led plan they could see 30 days out
The audits fed a 90-day content roadmap, equal parts diagnosis and strategy. We ran a competitive content gap analysis, pulled GSC data for a content audit (what to refresh, and where competitors were winning and Gladia was absent) and a keyword gap analysis across both the SERP and AEO sides, so the plan targeted commercial-intent search demand and AI answer gaps together.
We view topics through the lens of pillars and clusters. A pillar is a top-level topic like speech-to-text. A cluster sits beneath it, like real-time transcription. A fully built-out pillar normally spans a money page to capture commercial intent, a pillar page to capture informational intent and spokes to capture the longer tail of queries, all tied together with deliberate internal linking. For Gladia we mapped the pillars and clusters they needed to own and sequenced the work across 30, 60 and 90 days. The Voice Agents pillar went on to drive the biggest citation gains.
Gladia wanted us to work alongside their existing content team, so we worked inside a shared 30-day calendar to keep both teams from overlapping.
3. Content engineering: AI-assisted, human-owned content built to be cited
Most of what makes the content work happens before and around the writing. We've built our own content engineering platform: AI does the heavy lifting at each step, and our team owns the strategy, the judgment and the final word. Every piece runs through the same engineered pipeline:
- Evidence-grounded briefs. When Google and the AI engines are rewarding certain pages, we want to understand why. Each brief is built from that read of the live SERP, the People Also Ask questions and a map of where AI answers are won or lost.
- Built for retrieval. Each piece is structured on our CITABLE framework so a model can lift a self-contained, trustworthy passage and cite it, with relevant proof from sources like YouTube, Reddit and G2 built in.
- Quality gates that scale. Every claim is fact-checked against live sources, every link is validated and each piece is checked for cannibalization and duplication before it moves forward.
- Brand DNA enforced throughout. Gladia's persona, writing guide and style guide are applied at every step, so the voice holds across everything we ship.
- Schema, metadata and publish. We generate the structured data and metadata and publish straight into Gladia's CMS, so nothing lands back on their team's plate.
The result is content that holds up with an audience of engineers, who spot a weak claim immediately.
“There are very few agencies in this space that could help us with the kind of technical content that we have. It's not just B2B, it's also developer marketing, a niche within a niche.”
Anna, Head of Marketing4. Off-page: earning the Reddit citations AI models trust most
AI models weigh what others say about a company across the web, and Reddit is one of the highest-trust sources they pull from. Our own research into Reddit's influence on AI search found it appears in roughly 27% of the sources ChatGPT pulls from, which is why we invest there as heavily as we do. The AI visibility audit surfaces the exact threads models cite for Gladia's topics, which become a target subreddit map. The Reddit work then runs on the same pillars, clusters and focus keywords as the on-page content, so the two pull in the same direction. Link building runs in parallel from day one.
5. Measurement: reporting that ties AI visibility to pipeline
We treat AI visibility as the leading indicator. Pipeline is what it has to turn into. Everything is measured against a domain-specific set of buyer-language prompts, tracked at the level of each piece, each topic and Gladia overall, and reported across three layers:
- Leading indicators. Mention rate, citation rate and share of voice on the topics Gladia wants to own, alongside Google rankings for the commercial queries that actually drive revenue.
- Page-level engagement. Referral traffic and click growth, split by branded and non-branded, so we can tell genuinely new demand from people who already knew the name.
- Conversion. The most reliable read is self-reported attribution. When a prospect tells sales they found Gladia through an AI assistant, that is the signal we trust most.
AI search makes attribution harder than classic SEO, and we are upfront about that. A buyer often gets their answer inside ChatGPT or Claude without ever clicking, then arrives weeks later through a branded or direct search, so last-click reporting quietly credits “direct” for work AI search did. It is why we lean on self-reported attribution, and why it counts that Gladia's own team traces their sales-accepted lead growth back to LLM search.
Reporting runs through Gladia's HubSpot CRM, and Gladia sees the same live dashboard we work from, including how AI assistants describe them through our AI Perception module. The rhythm is deliberate: an async update every week, a live call every two weeks and a monthly review focused on business impact. The plan runs on a 90-day to monthly to weekly loop, so priorities can shift month to month as Gladia's do.
7x sales-accepted leads in four months, driven by AI search
Content was live within the first few weeks. Within four months, Gladia's monthly sales-accepted leads grew sevenfold and were still climbing into the following month.
A sales-accepted lead at Gladia is a prospect a rep has manually reviewed and judged to have real commercial potential. It is the metric the team uses to gauge pipeline health. AI search became the largest new driver of that growth, with the existing SEO work still pulling underneath it, and lead quality rose alongside volume.
“The quality of the leads at the sales-accepted lead level has also increased, with good and very good leads becoming more predominant in the overall mix.”
Anna, Head of Marketing7x growth in sales-accepted leads
Monthly sales-accepted leads grew sevenfold in four months and were still climbing into May. These are leads a rep manually judged to have real commercial potential.
AI search became the biggest new driver
Leads sourced from AI search grew roughly eightfold over the four months and, by April, made up 93% of all AI-referred sales-accepted leads.
Quality rose alongside volume
High-quality “good” and “very good” leads grew roughly threefold. The channel produced more qualified pipeline at a higher bar.
9x lift in LLM citation rate
Gladia’s citation rate in LLM answers rose ninefold, with the strongest gains in the Voice Agents topic. Citation rate climbed before the leads did.
