Perplexity spent late August wiring its Computer agent into the databases that financial analysts actually pay for. The company announced that Computer now connects to more than 20 new licensed finance data sources, including Dun & Bradstreet, Guidepoint, and IBISWorld, with no separate logins and every figure traced back to its source record. Perplexity finance data connectors sound like a niche analyst story. Treat them instead as the clearest evidence yet that AI search visibility is splitting into two channels, and only one of them can be won with content.
What Perplexity Actually Shipped
The announcement came from CEO Aravind Srinivas in late August: Computer, Perplexity's general-purpose work agent, now answers questions using a firm's own licensed data sources directly. An analyst asks a question. Computer assembles the answer from Dun & Bradstreet firmographics, Guidepoint expert calls, IBISWorld industry reports, and 20-plus other providers, without anyone juggling API keys or separate logins. Each number in the output links back to the record it came from.
We are doubling down on making Computer more useful for financial researchers and analysts. Computer now connects to new licensed data sources, including Dun & Bradstreet, Guidepoint, IBISWorld, and 20+ others. An analyst can ask Computer a question and get an answer built from https://t.co/6zDrHwNUYO
Aravind Srinivas@AravSrinivasAug 26, 2026
This wasn't a one-off. Perplexity launched Computer for Professional Finance back in May with licensed data from Morningstar, PitchBook, Daloopa, and Carbon Arc, plus 35 prebuilt finance workflows. A month later its Finance Search hit the Agent API, letting developers pull licensed datasets and live market data alongside cited web sources in a single tool call. By late July, FactSet fundamentals and expert-call transcripts were flowing in through connectors too. August's batch of 20-plus sources is the doubling down, and the pattern is unmistakable: the answer-engine changelog keeps adding sources you cannot publish your way into.
Why AI Engines Want Licensed Data
Finance is the domain where wrong answers carry liability, so it's the natural first market. To be honest, the logic applies anywhere money changes hands on the strength of an answer.
Open-web content has a provenance problem. Anyone can publish anything, which is why engines burn so much effort scoring trust signals. Licensed data solves provenance by contract: the provider maintains the records and stakes its business on their accuracy. The engine gets a clean citation trail as part of the deal. Perplexity has leaned into exactly that pitch, reporting benchmark results where its finance stack delivered the most accurate live financial data in its cohort at the lowest cost per correct answer.
There's a compounding pull here too. Professional users pay for accuracy. Accuracy improves with licensed sources. Licensed sources cost money, which subscription revenue funds. Expect the flywheel to reach legal, healthcare, and procurement data next.
Your competitors are getting cited by AI. You're not.
Every day without citation tracking is a day your competitors pull ahead in ChatGPT, Perplexity, and Claude.
The Two Channels of AI Visibility
Before I lose you, quick sidebar: this is the part that changes strategy, so it's worth slowing down for.
Until now, nearly everything written about AI search visibility assumed one channel. Be crawlable, be structured, be cited, be mentioned. That channel is real and still growing. But answers about companies and markets are increasingly assembled from a second channel that starts behind a login.
| Open-retrieval channel | Licensed-data channel | |
|---|---|---|
| How the engine gets it | Crawling and search over the public web | Commercial contracts with data providers |
| Who controls entry | Anyone who publishes | The provider and the platform |
| What decides your presence | Content structure, authority, citations | What the provider's records say about you |
| Your lever | Publish and earn mentions | Maintain records, earn database presence |
Look at the fourth row twice. When a buyer asks an AI agent to size your market or shortlist vendors, part of that answer now comes from records you've probably never audited.
What This Means for B2B SaaS Teams
Your company already lives inside this channel whether you participate or not. Dun & Bradstreet keeps a firmographic record on your business. IBISWorld writes the industry reports that describe your category and name its players. Guidepoint-style expert networks record calls where practitioners describe the tools they use, in their own words, to buyers doing diligence. And PitchBook-class databases already carry your funding history and investor list.
So the response splits into four jobs.
- Audit your firmographic records. Pull your D&B profile and fix stale revenue bands, headcount, category codes, and addresses. A buyer's agent citing a wrong record is a lost deal you'll never see happen.
- Chase industry-report presence. Analysts license IBISWorld and its peers. Being a named player in your category's reports is now AI answer material, and briefing the analysts who write them is a visibility activity, not corporate vanity.
- Respect the expert-call channel. When your customers talk to expert networks, their words become retrievable research. The fix is unglamorous: a product people describe accurately and enthusiastically without coaching. Trust signals compound differently when money is involved, and this channel is where they compound hardest.
- Keep winning the channel you control. The open web still decides most citations for most queries, and getting your content into the sources AI engines actually pull from remains the highest-return work. Structured, citable content on your own domain feeds both retrieval and, eventually, the datasets these providers build.
Which raises the obvious question: how would you even know what the licensed channel says about you? You mostly can't read it directly. What you can do is watch the output side. Tracking how AI engines describe your brand across engines, weekly, is the only practical alarm for a wrong revenue band or a stale category label surfacing in answers. Record errors get fixed in an afternoon. Noticing the next one before your buyers do is the ongoing job.
Budget reality: the D&B audit is an afternoon, analyst outreach is a quarterly habit, and answer monitoring is a weekly check. Or RankControl's agents run the monitoring side continuously and flag description drift while you handle the records only a human can fix.
200+ SaaS teams already track their AI citations.
They know exactly when ChatGPT mentions their brand, and when it stops. Do you?
Where This Goes Next
Watch four things this quarter. Whether Perplexity extends connectors past finance into legal and healthcare data. Whether OpenAI and Google answer with their own licensing sprees, the way both already license news. Whether B2B data providers start marketing "AI answer presence" to the companies in their databases, which would make the licensed channel pay-to-describe in the ugliest way. And whether buyers begin expecting vendor claims to trace to a licensed record before they'll repeat them internally.
For the record, the durable read is simpler: engines are converging on answers that can prove where every number came from. Provenance is becoming the product. The companies that treat their data trail, from their own site to third-party records, as a single visibility surface are the ones AI answers will keep describing correctly.
How often does ChatGPT mention your brand?
Most founders have no idea. The answer might surprise you.
