Disconnected information
Bring the sources and dates behind an answer into the workflow that produces the page.
Search-led implementation
One page family. One AI-assisted workflow. One agreed improvement.
Targeted Impressions is James Gibbons’s independent practice and lab for search-led website improvements and AI-assisted publishing. The work connects opportunity analysis, implementation, and checks on what customers actually receive.
Start with the problem
Your source data, page copy, and software can move at different speeds. Start where that mismatch affects the work.
Bring the sources and dates behind an answer into the workflow that produces the page.
Check whether the wording preserves the facts, qualifications, and limits of the source.
Inspect what reached the customer instead of treating a completed task as proof of delivery.
Give the work a clear owner, bounded permissions, and a usable handoff.
A bounded engagement
Scope, access, and acceptance are agreed before work. No guaranteed rankings, AI citations, or revenue lift.
01
Inspect what exists, the available evidence, and what cannot yet be measured.
02
Agree one problem, one useful outcome, and how to judge the result.
03
Make the authorized change, or deliver an implementation-ready package when access is outside the engagement.
04
Inspect the delivered output and test the relevant failure cases.
05
Leave clear ownership, a recovery path, and a proportionate measurement plan.
The lab
These projects support the practice with examples and experiments. A working demonstration does not establish client adoption or guaranteed results. Each project has its own scope and limitations.
Independent project
Flight intelligence as a listening surface instead of another booking interface.
Independent project
National-park discovery and listening. Modeled previews are not live field observations.
Independent project
An exploration of sound as an interface for marketing information.
Independent project
Checks and publishing rules for websites updated with AI. Inspect each example to see what it actually tests.
Independent project
Visual and optional audio cues for system events. Rendering an event does not verify its source claim.
Interface experiment
The example below is authored and browser-local, not live monitoring. Sound begins only when you choose it. It does not establish an improvement in attention or productivity.
A composed tone pattern accompanies example messages. It does not detect or verify an event.
Authored examples, not production events
Optional sound; no autoplay
Text remains available without audio
Start a conversation
Tell me what your team does today, where it gets stuck, and what a useful result would look like. Start with a description—not credentials or confidential customer material.
Prefer email? [email protected]