In just 69 days, an established B2B tech business became more accurately represented and more frequently recommended in AI answers.
The challenge
The company had an established business, satisfied customers, and a simple question: when potential customers asked AI which providers to consider, was it making the list?
The answer was mixed. Sometimes it appeared. Sometimes competitors were recommended instead. But the review uncovered something more concerning: some AI answers confused the company with another business.
The problem went beyond being overlooked. AI did not consistently understand who the company was.
The approach
The engagement began by examining what potential customers could encounter across five AI engines.
Using a fixed set of 521 questions, the review assessed 1,143 answers. It looked beyond whether the company appeared: was the information accurate, was its content used as evidence, and was it recommended when someone asked which provider to choose?
Establish how AI describes and recommends the company.
Address the information gaps revealed by the assessment.
Repeat the same question set and compare what changed.
The findings pointed to three priorities: clarify the company’s identity, make its website content easier to understand, and strengthen the evidence available through external sources.
What changed
The work began with the information AI needed to understand the business correctly, then expanded to the sources that could support its recommendations.
Before
The website did not clearly state the company’s legal name or independence.
What changed
The legal name was added, the absence of a parent organization was stated explicitly, and the website was linked to the company’s external profiles.
Before
Different pages shared the same description, making their individual purposes harder to distinguish.
What changed
Page-specific descriptions helped distinguish service information, reviews, and answers to customer questions.
Before
An important third-party profile was unclaimed, another was missing, and more content was needed to answer comparison questions.
What changed
One profile was claimed, another was created, and company information was made consistent across those sources. Comparison content was also added to help answer questions about which provider to choose.
With those changes in place and further content work still underway, the assessment was repeated 69 days after the baseline.
Some AI answers were associating the business with another company. Clarifying its identity took just two hours, and addressed a problem that more brand building could have left untouched.
That’s why early results matter. When the assessment finds a specific, fixable gap, progress can begin quickly. The next measurement, 69 days after the baseline, showed how much had changed.
The results
In less than ten weeks, the company was understood more accurately, cited more frequently, and recommended more often.
The company earned consideration more often when AI was asked which providers to choose.
Across 100 assessed ranking and shortlist cells.
AI used the company’s own information more frequently to support its answers.
Among runs where citations were observable.
Far fewer answers misrepresented the business or pointed to another company.
Across 1,143 assessed answers.
For the company, this meant a stronger presence when potential customers were exploring their options: fewer misleading answers, more use of its own information, and more frequent recommendations.
After just 69 days, progress was measurable. The company had a clearer AI presence and a baseline for tracking what improved next.
These results measure changes in AI answers. Effects on leads and revenue were not established. Results and timing are specific to this engagement.
The review used a fixed set of 521 questions across five AI engines, with 1,143 answers assessed. Re-measurement took place 69 days after the baseline using the same question set. Recommendations were assessed across 100 ranking and shortlist cells. Citation rates cover runs where citations were observable. Problematic answers were counted across 1,143 answers. These measures have different denominators and should be read separately.
Find out whether your company appears, how accurately it is described, and which competitors are being recommended.