Media research, audit reports and PR outreach preparation

In the retrospective account, XWECAN generated audit reports for prospective clients who wanted a review of their website or marketing resources. The pipeline called research tools including Semrush and Ahrefs, looked at Google Trends and collected client website information. When a source did not offer a suitable API, Noah describes using Playwright to obtain the relevant material through the browser.

The research and writing stages had different jobs. Smaller models extracted specific information, such as useful material from a website’s HTML; larger models composed the final report. The account does not specify model versions or measured report-cost savings.

The PR problem was more specific than obtaining a large media list. Noah describes a conventional approach of buying or finding lists for a broad topic and sending the same story widely. His alternative began with the research a person would do to decide whether an individual journalist might actually care about a story.

Published articles supplied the context. The retrospective account describes collecting articles from major publications and using small models to derive the author’s interests within a broader domain. A technology label alone was too coarse. The system looked for the narrower theme that linked the client’s story to something the journalist had already explored.

Another recording describes generating queries for those narrow subjects and incorporating recent coverage into outreach preparation. The value of that workflow was the connection between the research and the pitch: the story angle came from a defensible fit with the journalist’s coverage, rather than from a generic category assigned to the whole publication.

A self-authored application draft gives additional retrospective detail about the research flow. It describes taking a client summary and the subject seeking press, consulting prepared tables of major publications by industry and generating queries to discover relevant articles and journalists. That makes the story brief an explicit input: the search was intended to find evidence of interest in that particular subject.

The same draft describes downloading articles, extracting author names and contact information, and cross-referencing an existing journalist email database with previous delivery results. It then describes scoring the journalist’s relevance and adapting the subject and body of a prepared email sequence to the research context. These details help explain the intended record relationships and how research informed outreach preparation.

This retrospective account complements the walkthroughs; it is not a new execution test. Exact production configuration, query volumes and measured outreach outcomes have not been reconstructed.

The outreach account also describes considering the journalist’s relevance, the potential value to the client of a placement and the capacity available on a channel. Programmatic filters and AI helped choose whether a contact opportunity justified a message or email. Contact information was sourced and the system routed outreach to an appropriate platform.

These retrospective descriptions explain acquisition, extraction, article and author context, relevance assessment and output. They do not establish the complete configuration, scoring thresholds or final approval workflow.

Noah separately confirms that he set up new outreach systems and that the team secured the authorized Fujitsu CNN coverage shortly afterward. That makes the system relevant to the broader operating story, but it does not prove that one matching rule or generated pitch caused the placement.

Noah set up the research and outreach systems. The team delivered the press coverage. The recordings explain the workflow and its decision logic without establishing a numerical response, cost or placement improvement.