How to Use the Sellm API for LLM Visibility to Build Your Industry Reports

AI search has quietly become a discovery channel. When someone asks ChatGPT for the "best 65-inch TV for a bright living room" or asks Perplexity which CRM fits a small sales team, the model returns a short list of brands. That list is the new shelf. If your brand is on it, you win consideration. If it is not, you are invisible, and you usually have no idea it happened.

The problem is that this shelf is invisible to every tool marketers already use. Google Search Console shows you keyword rankings. Your analytics show you sessions. Neither tells you how ChatGPT, Claude, Perplexity, Gemini, Grok, or Copilot describe and rank your brand when a buyer asks a real question. That gap is exactly what an LLM visibility API is built to close, and it is why we built the Sellm API.

This article walks through both halves of the story: the business case, using a real industry report we published, and the technical configuration, so your team can build the same thing for your own market.

Why an API, not just a dashboard

Plenty of teams want AI visibility data, but they do not all want the same shape of it. An agency wants to drop numbers into client decks. A growth team wants visibility KPIs in their own BI stack next to acquisition and revenue. A data team wants to run a one-off industry study without onboarding a whole platform.

A dashboard cannot serve all of those. An API for LLM visibility can. With programmatic access, any team can pull the underlying data and build their own custom reports, dashboards, and alerts, on their own terms. You are not limited to the views a vendor decided to ship. You get the structured signal and you decide what to do with it.

That is the core promise of an AI visibility monitoring API: the same data that powers a polished dashboard, exposed as clean JSON you can build on.

A worked example: the Samsung TV report

The best way to understand what is possible is to look at something we actually built. We published a public study of how AI search recommends televisions, and you can explore it at sellm.io/samsung. Here is the methodology, because it is the blueprint you would reuse for any industry.

1. We designed a prompt matrix

Instead of guessing a handful of queries, we built a matrix of realistic buyer questions by combining dimensions: product type, use case, screen size, and price range. "Best TV for gaming under 55 inches," "best OLED TV for a bright room," and so on. This matters because LLM recommendations are query-sensitive. The brands that win "budget TV" are not the brands that win "premium home cinema." A good industry report covers the real spread of intent, not one lucky query.

2. We ran every prompt with replicates

This is the part most people miss. LLM answers are non-deterministic: ask the same question twice and you can get different brands in a different order. A single run is an anecdote, not data. So for each prompt we executed multiple searches and aggregated the outcomes. That turns "ChatGPT mentioned Samsung once" into "Samsung appears in 73% of responses for this query, at an average position of 2.1." Statistical significance is the difference between a screenshot and a study.

3. We extracted structured signal from each answer

For every response we pulled out the brands mentioned and their order, the specific products named (down to model level, like the Samsung QN90F), and the sources the model cited to ground its answer. That last point is the actionable one: if a handful of review sites are driving most recommendations in your category, you now know where to focus.

4. We aggregated it into a report

Searches analyzed, brand mentions, product mentions, share of voice, and cited domains, sliced by query and by AI engine. The result is a clear picture of who owns the AI shelf in the TV category and why. None of this is specific to televisions. Swap the prompt matrix and the brand, and the same pipeline produces a credit-card report, a project-management-software report, or a report on your own category.

How to configure the API to build it yourself

Now the technical half. Here is how to go from zero to a working industry report with the Sellm LLM visibility API.

1. Get an organization key and credits

Create an organization API key from your Organization API Integrations page, then buy prepaid credits. Pricing is pay-as-you-go: credits cost $0.005 each, with a $5 minimum, and no subscription. One credit covers one prompt, on one provider, in one location, for one replicate. So your total cost is simply:

credits = prompts × providers × locations × replicates

A 40-prompt study across 2 providers, 1 location, at 5 replicates is 400 credits, or $2.00. There is no per-seat fee and no minimum monthly spend, which positions Sellm as one of the most affordable AI visibility tools on the market. That combination of predictable, granular pricing is what makes broad, statistically solid studies realistic to run, not just once, but on a recurring schedule.

2. Submit an async analysis

The core endpoint is POST /v1/async-analysis. Organization keys pass a brandName on each request, so you can analyze any brand, including competitors or prospects, without setting up a project. A single call fans out across every engine you list, so you get ChatGPT API brand monitoring, Perplexity API brand monitoring, and the other major models in one normalized response, instead of building and maintaining a separate integration per provider. You send your prompt set, the engines you care about, the markets, and how many replicates to run.

curl -X POST https://sellm.io/api/v1/async-analysis \
  -H "Authorization: Bearer sellm_your_org_key" \
  -H "Content-Type: application/json" \
  -d '{
    "brandName": "Samsung",
    "prompts": [
      "best OLED TV for a bright living room",
      "best 65 inch TV for gaming under 1000"
    ],
    "providers": ["chatgpt", "perplexity"],
    "locations": ["US"],
    "replicates": 5,
    "webhook": { "url": "https://your-app.com/webhooks/sellm" }
  }'

Two configuration choices define the quality of your report. Prompts should mirror the buyer-intent matrix from the Samsung example, not a single keyword. Replicates are how you buy accuracy: more executions per prompt mean the aggregated share of voice and positions reflect real patterns rather than one sampling of a non-deterministic model. For a serious report, 5 to 10 replicates per prompt is a sensible floor.

3. Receive results by webhook or polling

The analysis runs asynchronously. You can either configure a webhook to receive the payload when processing completes (deliveries are signed with HMAC-SHA256 and retried for up to 24 hours), or poll GET /v1/async-analysis/{analysisId} until status is succeeded. Webhooks are the better fit for a pipeline; polling is fine for a quick script.

4. Read the structured response

Every analysis returns the same normalized schema whether you poll or use a webhook:

{
  "data": {
    "id": "aa_01JXYZ...",
    "status": "succeeded",
    "brandName": "Samsung",
    "summary": {
      "sovPct": 34,
      "coveragePct": 82,
      "avgPos": 2.1,
      "sentiment": 0.71
    },
    "results": [
      {
        "prompt": "best OLED TV for a bright living room",
        "provider": "chatgpt",
        "position": 2,
        "brandsMentioned": ["Samsung", "LG", "Sony"],
        "citedDomains": ["rtings.com", "samsung.com"]
      }
    ]
  }
}

The summary block gives you the headline KPIs: share of voice (sovPct), how often you appear at all (coveragePct), average rank when mentioned (avgPos), and overall sentiment. The results array gives you the per-prompt, per-provider detail, including the full ordered brandsMentioned list (your competitive landscape) and the citedDomains driving each answer.

5. Build the report

From there, the report is yours to shape. The summary numbers become your top-line cards. Iterating results and counting brandsMentioned across the full set reconstructs the competitive share-of-voice chart. Aggregating citedDomains produces the "who is influencing the AI" view that makes the data actionable. Push it into a BI tool, a client deck, a Slack alert when your share drops, or a public study like the Samsung one. Because the output is consistent JSON, the same code works for any brand and any industry you point it at.

Start building

AI search is where a growing share of buying decisions begin, and right now most brands are flying blind on it. An AI visibility monitoring API turns that blind spot into a measurable, repeatable report. We built the Samsung TV study on exactly this pipeline, and the same endpoints are available to your team today.

Grab an organization key, buy $5 of credits, and run your first analysis in minutes. See the full reference on the LLM Monitoring API page, or explore the related LLM Mentions API for tracking brand mentions across engines.