New ChatGPT recommendations for an established injury firm.
An established personal injury firm engaged Flaregraph to improve how AI understood its expertise. The work began with technical diagnosis and schema corrections, followed by service and location content. Flaregraph prepared specifications and briefs for the firm’s team to implement, then verified the work as it went live. The engagement is ongoing.
Existing expertise. Missing recommendations.
The firm represents people injured in accidents and at work. The question was whether AI assistants connected that expertise to the situations in which prospective clients needed it.
Flaregraph recorded the answers to twelve fixed questions about legal representation, then returned to those questions in July, August, and September 2026.
From diagnosis to implementation.
Flaregraph began with the website’s schema: the structured data that identifies the firm, its services, and the places it serves. Technical specifications addressed identity links, outdated review data, service definitions, and geographic coverage.
The work also addressed redirects and service-page technical details, alongside content briefs for construction accidents and claims an injured worker might bring beyond workers’ compensation. Research, engineering, and content informed the same implementation plan.
The firm’s team implemented the changes. Flaregraph verified the published work and reviewed how the answers changed.
The firm began appearing in answers to two relevant questions.
Neither question returned a ChatGPT recommendation for the firm in the July or August samples. In September, both returned the firm in every one of the three answers collected.
| Question topic | July | August | September |
|---|---|---|---|
| Finding a construction-accident lawyer | 0 of 3 | 0 of 3 | 3 of 3 |
| Finding a firm for a claim beyond workers’ compensation | 0 of 3 | 0 of 3 | 3 of 3 |
Question topics are paraphrased, with location details withheld. Results reflect the same fixed, market-specific questions across each review.
The complete record.
Each review collected three answers to each of twelve questions across ChatGPT, Claude, Gemini, and Perplexity: 144 answers in total. Qualifying appearances included recommendations and unsponsored local-listing appearances, not paid advertisements.
| Assistant | July 13–14 | August 7 | September 1–2 |
|---|---|---|---|
| ChatGPT | 5 of 36 | 0 of 36 | 11 of 36 |
| Claude | 4 of 36 | 4 of 36 | 5 of 36 |
| Gemini | 0 of 36 | 1 of 36 | 0 of 36 |
| Perplexity | 0 of 36 | 0 of 36 | 0 of 36 |
| Total | 9 of 144 | 5 of 144 | 16 of 144 |
The firm appeared on four of the twelve questions in July, three in August, and eight in September. A question counts here when at least one collected answer returns the firm.
The September improvement was concentrated in ChatGPT, with a smaller increase in Claude. Gemini and Perplexity did not return qualifying appearances in that review.
What the change means.
The firm became visible in specific recommendation answers where it had previously been absent. The pattern is consistent with the work contributing to that change, although the observations do not isolate its cause.
This is a dated record of sampled AI answers—not a count of enquiries, signed cases, or revenue. The same questions can be reviewed again to see whether the appearances persist.
The next phase of work.
This is an ongoing engagement. The initial work strengthened the firm’s technical foundation and service information, with new recommendations appearing in ChatGPT and established local visibility continuing in Claude.
The next phase focuses on extending that visibility to additional questions and to Gemini and Perplexity, while continuing to measure progress across all four assistants.