Executive Summary

When buyers ask AI platforms like ChatGPT, Claude, or Perplexity for product recommendations, conventional SEO reporting falls blind. Blesso helps B2B SaaS companies track, diagnose, and capture visibility inside AI search responses.

With Claude, Blesso

• Lifted client citation share across tracked buyer prompts from 12% to 31%
• Diagnoses why a competitor is cited where the customer is not, and turns that into a prioritised content plan rather than a rank report
• Drafts LinkedIn content against a per-customer voice profile, with mandatory human approval before anything publishes
• Cut the time from identifying a gap to publishing against it from 14 days to 4 days

The Challenge

A buyer asks an LLM which tools to consider. Three names come back.

If you are not one of those three, you were not rejected. You were never in the evaluation at all.

For B2B SaaS, the first pass of the buying process has quietly moved. A buyer who would once have opened Google and clicked through four vendors now asks an LLM to tell them which tools solve the problem and how they compare. What comes back is a synthesised answer naming a handful of products. There is no page two, and no impression to win. You are in the answer or you are absent from it. Agentic coding has sharpened this by making it far easier to ship a product than to become known for one.

The difficulty is that this is hard to even see. Conventional rank tracking reports that a page sits at position four for a keyword. It cannot tell you whether an LLM named you when a buyer asked a question in their own words, and every buyer phrases it differently.

Blesso engaged Searce, an Anthropic partner, to build the system that closes that gap.

The Solution

Two jobs for Claude

To address this gap, Blesso partnered with Searce to build a system powered by Claude that continuously tracks brand presence across answer engines. Rather than simply displaying raw data on a dashboard, Claude performs two critical, high-judgment tasks:

Working out why a competitor wins the citation

Claude reads what the answer engines surfaced and the material they drew on, and identifies why a competitor is being named where the customer is not, whether it is a specific claim the competitor has published, or a category of source the customer has no presence in. The output is a prioritised set of content to produce, each item tied to the question it is meant to win.

The reasoning behind building it this way is that a ranking gap and a citation gap are different problems. The keyword you rank poorly for may be costing you nothing. The question where the model had nothing of yours available to cite is the one that loses the deal, and it frequently corresponds to no keyword the customer was tracking.

Drafting in the customer's voice, published only by a human

Claude drafts against a voice profile built from each customer's own writing, so output reads like the founder rather than like an agency. Approval is mandatory rather than optional: every draft routes back to the customer for review, and nothing publishes unreviewed. That is a deliberate constraint. The asset being built is authority with a specific audience, and one post that sounds wrong costs more credibility than ten adequate ones earn.

Why Claude

The two jobs are model-sensitive in opposite ways. The diagnosis work is inference rather than retrieval: identifying why an answer engine reached for a competitor's material, rather than simply noting that it did, means reasoning over a long and messy corpus. The voice work demands the reverse, applying the same voice profile consistently across hundreds of drafts without drifting toward a generic register, which is the characteristic failure of weaker models on long-running generation.

The platform is deliberately model-agnostic and can run on other providers. Claude was selected for the customer-facing analysis and generation path, where both of those properties carry the load.

The Outcome

Cited where the evaluation actually happens

• Increased citation share: Median client citation share across tracked buyer prompts grew from 12% to 31%.
• Faster execution: Time required to identify a competitive citation gap and publish targeted content dropped from 14 days to 4 days.
• Organic traffic growth: Median monthly non-branded organic traffic increased from 10,000 to 15,000 visits per customer.
• Scale to date: 15 active enterprise clients with 180 human-approved content assets published.