An entity resolution and identity data platform started near the bottom of its tracked category. In the first agreed audit run on 26 May 2026, it appeared in 3.5% of relevant AI answers and ranked tenth among eleven companies. By the latest completed weekly window on 10 August, visibility had reached 23.2%. Its current trailing category rank was first.
TL;DR
- AI visibility rose from 3.5% to 23.2%, a 6.7 times increase.
- Share of voice rose from 4.4% to 22.1%, a 5.1 times increase.
- Category rank improved from tenth to first.
- Public crawl access, owned content and independent sources drove the organic result.
- AI-assistant referral visitors rose from 12 to 64 across the measured periods.
| Metric | Value | Caption |
|---|---|---|
| AI visibility | 3.5% to 23.2% | |
| Share of voice | 4.4% to 22.1% | |
| Category rank | 10th to 1st |
What was the starting position?
In the first agreed run on 26 May 2026, the platform held 3.5% visibility, 4.4% share of voice and tenth place in the eleven-company set. It had strong expertise, but it was absent from too many technical and buying questions.
The number on its own was not the useful part. The pattern behind it was. Where the platform's content was public and indexable, it already won. It held one of the strongest average positions in the whole category on the bottom-funnel prompts about scaling entity resolution to hundreds of millions of records in real time. But its most detailed explanatory content on AI-native identity resolution, the material written to answer exactly the questions a technical buyer researching this category types into an AI engine, sat behind a login gate. AI engines cannot crawl a page behind authentication. The platform was close to invisible on an entire growing subtopic, agentic AI and MCP-era entity resolution, not because the content was thin, but because it was unreachable.
Who else is answering these questions?
The neutral competitor set included AWS Entity Resolution, Data Ladder, Informatica, Quantexa, Reltio, Semarchy, Senzing, Splink, Tamr, Tilores and Zingg. The list is alphabetical and does not identify which company this study covers.
What did we actually do?
Nothing here was a single lever. It was closing an access defect, rescuing content that was already half-working, and then systematically filling the gaps a straightforward audit surfaced.
Fixed the crawl-access defect first. The platform's flagship identity-resolution explainer content was sitting behind a login wall. We moved it onto public, indexable pages, so AI crawlers could reach it for the first time. This was the single change with the widest blast radius: it is hard to be cited for content an engine has never been able to read.
Rescued content that was already half-working. Existing articles appeared in retrieval but lost the citation to competitors. We rebuilt them with direct answers, comparison blocks, useful lists and clearer schema.
Built content around measured gaps. The programme covered technical explainers, category pages, buyer guides, comparisons, alternative pages and practical how-to articles. Each page answered a question where the company had lacked a useful public source.
Matched the format the engines already reuse. A rival's eight-vendor comparison matrix was the only page being cited for a specific shortlist prompt. We rebuilt a comparison page in the same extractable format, with the same structural elements AI engines already pull from that competitor's page, so ours had an equal shot at the citation.
Built a wider public source base. Editorial placements added independent support. Medium essays, LinkedIn posts and YouTube carried the same ideas onto other useful surfaces. Reddit and community work focused on answering real questions, not posting links.
Kept a social cadence running. Regular LinkedIn posts pointed back to the new material, which does not move an AI engine's index directly but does build the kind of discoverable trail engines increasingly weight.
What actually moved, and by how much?
| Metric | First audit run | Latest weekly window |
|---|---|---|
| AI visibility | 3.5% | 23.2% |
| Share of voice | 4.4% | 22.1% |
| Category rank | 10th | 1st |
Visibility grew 6.7 times and share of voice grew 5.1 times. The company moved from near the bottom of the tracked set to number one.
Did the impact show up in website and agent traffic?
AI-assistant referral visitors rose from 12 to 64 across the two measured web periods, a 209% increase in the daily rate. Edge logs also recorded 41,034 identified AI-agent requests from 13 July onward. There is no earlier like-for-like baseline for the agent-request figure because that log collection began partway through the programme.
How should the numbers be read?
AI visibility is the share of tracked answers that name the company. Share of voice is its share of all tracked brand mentions. Category rank compares visibility across the tracked companies. The current rank uses the completed trailing 30-day view, which is more stable than one day.
The prompt set and daily monitoring matured during the engagement. These figures show the full programme direction from first audit to the latest completed week. They are not presented as a laboratory test with a frozen denominator.
What is the practical lesson?
The result did not come from publishing the most material. It came from fixing what engines could not read, then covering the technical, comparison and buying questions where the company had no credible public answer. Owned blog content was the base. Editorial, Medium, LinkedIn, YouTube and community sources reinforced the same story.
If you want to see where your own category stands, the free AI search audit shows your current visibility, share of voice and competitive position.
Squarespace
Webflow
Sanity
Shopify
HubSpot