02 / THE REPORT
The view by buyer need.
Answers naming your brand versus the most-named option in each group.
Buyer needFieldnoteMost named option
team collaboration83 / 120DDaybridge96 / 120 answers
project planning65 / 120DDaybridge96 / 120 answers
resource management44 / 120DDaybridge96 / 120 answers
reporting36 / 120DDaybridge96 / 120 answers
enterprise controls17 / 120DDaybridge96 / 120 answers
Additional observations
Findings from inspected evidence
01Clarify who can approve changes and see project historyYour website · Visibility effect untested · Product marketing + content · Half-day brief; publishing varies
02Publish a practical guide to client reportingUseful content · Visibility effect untested · Content lead + subject-matter expert · Half-day brief; research and publishing vary
01 / Product marketing + content · Half-day brief; publishing varies
Clarify who can approve changes and see project history
WHAT YOU’RE WORKING TOWARDS
A verified permissions table on the existing features page, covering roles, approvals and access to project history.
THE FINDING
Claude recommended Daybridge in 24 of 30 answers about “enterprise controls” and Fieldnote in 1 of the same answers (gap 77 points; plausible range 56–87). The gap holds in both ways we asked and in 3 of 3 repeats. Gemini recommended Daybridge in 24 of 30 answers about “enterprise controls” and Fieldnote in 3 of the same answers (gap 70 points; plausible range 49–81). The gap holds in both ways we asked and in 3 of 3 repeats. The same kind of gap also passed our checks on 2 more engines.
Your own pages are the part you control directly.
Visibility effect untested. This report measures a gap; it does not establish that the proposed change would improve AI visibility.
Selected evidence E02E03
ALREADY COVERED
The illustrative features page already names enterprise controls and permissions.
THE REMAINING QUESTION
The excerpt does not explain which roles can approve changes or view project history. Check the full page before adding that detail.
WHY IT HELPS THE BUYER
A buyer can check whether the documented approval workflow fits their team before booking a demo.
1Ask the product owner to confirm which roles can approve changes and view project history.
2Check the full features page for an existing explanation before adding content.
3If the detail is missing, add a small permissions table with plan availability and a real product example. If it is already clear, leave the page unchanged.
DONE WHEN
A buyer can identify the available permissions and their plan requirements on the features page, with the product owner’s sign-off.
CHECK THE RESULT
Recheck how often Fieldnote is recommended for the same questions, no sooner than three weeks after the change is live. Claude 1 of 30 today; a repeat normally shows 0–6, so 7 or more would be a real change. Gemini 3 of 30 today; a repeat normally shows 0–9, so 10 or more would be a real change. OpenAI 4 of 30 today; a repeat normally shows 0–10, so 11 or more would be a real change. Perplexity 5 of 30 today; a repeat normally shows 1–12, so 13 or more would be a real change. No untouched buyer need is available as a control, so treat any change as inconclusive unless it is large. Smaller real gains may not show in this many answers.
Daybridge is recommended for its role-based permissions, approval controls and audit history for larger teams.
OpenAI E04
Interpretation. This fictional excerpt names permissions but does not describe roles. The task checks the full page and product facts before changing anything.
Published research. In controlled tests where the page was already available to the model, pages that matched the question and stated prices, specifications and a current date were cited first more often; formatting changes alone had no effect. https://arxiv.org/html/2605.25517v1
Published research. A 2026 review of 45 studies found no technique with a stable, cross-platform causal effect on whether AI engines find a page. https://arxiv.org/html/2607.14035v1
Treat this as an experiment: no controlled study shows that page edits reliably change AI recommendations.
02 / Content lead + subject-matter expert · Half-day brief; research and publishing vary
Publish a practical guide to client reporting
WHAT YOU’RE WORKING TOWARDS
A guide with a reusable client-report outline, an annotated example and guidance on choosing metrics. Update an existing resource if it already covers the question.
THE FINDING
Claude recommended Daybridge in 24 of 30 answers about “reporting” and Fieldnote in 4 of the same answers (gap 67 points; plausible range 45–79). The gap holds in both ways we asked and in 3 of 3 repeats. Gemini recommended Daybridge in 24 of 30 answers about “reporting” and Fieldnote in 6 of the same answers (gap 60 points; plausible range 39–73). The gap holds in both ways we asked and in 3 of 3 repeats. The same kind of gap also passed our checks on 2 more engines.
Your own pages are the part you control directly.
Visibility effect untested. This report measures a gap; it does not establish that the proposed change would improve AI visibility.
Selected evidence E05E06
ALREADY COVERED
The fictional features page lists reporting for teams but does not teach a client-reporting workflow in the inspected excerpt.
THE REMAINING QUESTION
Which metrics belong in a useful client report, and how should a team explain budget variance and next decisions? Check the resource library before commissioning a guide.
WHY IT HELPS THE BUYER
An agency marketer can prepare a client report from a reusable outline and understand where the existing product fits.
1Check existing articles, help pages and templates for client-reporting guidance before commissioning a new piece.
2Ask a client-services lead which project metrics clients actually need and how reporting cadence varies. Use a real, permission-cleared example.
3Build the guide around selecting metrics, explaining budget variance and recording next decisions. Include a usable report outline and the limits of the example.
4Show how the existing product supports the workflow only where verified. Link the finished resource from the relevant product and help pages.
DONE WHEN
A reader can prepare a useful client report from the guide and example. The article states its assumptions and links to verified product instructions.
CHECK THE RESULT
Recheck how often Fieldnote is recommended for the same questions, no sooner than three weeks after the change is live. Claude 4 of 30 today; a repeat normally shows 0–10, so 11 or more would be a real change. Gemini 6 of 30 today; a repeat normally shows 1–13, so 14 or more would be a real change. Perplexity 9 of 30 today; a repeat normally shows 3–16, so 17 or more would be a real change. OpenAI 9 of 30 today; a repeat normally shows 3–16, so 17 or more would be a real change. No untouched buyer need is available as a control, so treat any change as inconclusive unless it is large. Smaller real gains may not show in this many answers.
Daybridge is recommended for teams that need scheduled client reports and project-level budget visibility.
OpenAI E07
Interpretation. The fictional excerpt establishes the reporting topic, not the absence of an existing guide.
Published research. In controlled tests where the page was already available to the model, pages that matched the question and stated prices, specifications and a current date were cited first more often; formatting changes alone had no effect. https://arxiv.org/html/2605.25517v1
Published research. A 2026 review of 45 studies found no technique with a stable, cross-platform causal effect on whether AI engines find a page. https://arxiv.org/html/2607.14035v1
Treat this as an experiment: no controlled study shows that new content reliably changes AI recommendations.
MAKE THE NEXT CHECK COMPARABLEKeep your questions and market consistent. Make a change you can document, then repeat the measurement. A movement in mentions alone does not establish what caused it.
04 / THE REPORT
A bounded measurement. An open record.
This is directional evidence for competitive awareness, within a defined set of questions. It does not measure every AI conversation or predict sales.
Questions and original answers: English. Report commentary: English. Changing the interface language does not collect new answers or translate the saved evidence.
Market & languageUnited Kingdom · English
Categoryproject management software
Intended audiencegrowing teams
Collection window (UTC)Sep 19, 2026 – Sep 19, 2026
Provider configuration & coverage.
26 question groups · two phrasings · 3 repeats per provider. 624 of 624 successful answers analyzed.
Provider / modelAnalyzed / collectedWithout sources
OpenAIIllustrative fixture; no model was queried
156 / 1560 / 156GeminiIllustrative fixture; no model was queried
156 / 1560 / 156ClaudeIllustrative fixture; no model was queried
156 / 1560 / 156PerplexityIllustrative fixture; no model was queried
156 / 1560 / 156Wording & question fit.
Brand mentions for matched question / repeat pairs. Differences include ordinary response variation; these are not confidence intervals.
OpenAIFirst phrasing: 35/78. Second phrasing: 35/78. Difference: 0.0 percentage points. 78/78 matched pairs.
GeminiFirst phrasing: 30/78. Second phrasing: 31/78. Difference: 1.3 percentage points. 78/78 matched pairs.
ClaudeFirst phrasing: 25/78. Second phrasing: 25/78. Difference: 0.0 percentage points. 78/78 matched pairs.
PerplexityFirst phrasing: 36/78. Second phrasing: 36/78. Difference: 0.0 percentage points. 78/78 matched pairs.
On topic: 0 · Mixed: 0 · Off topic: 0 · Unclear: 0 · Not assessed: 624. Topic labels are machine assessments. No answer is removed from the overall counts on this basis. Not assessed does not mean passed.
Buyer-need questions: 245/600 answers naming your brand · Category-only check: 8/24 answers naming your brand. Different scopes, not a controlled estimate of bias.
Protocol: buyer-panel-v2. 52 distinct prompt texts. 624 saved prompts match the approved questions; 0 differ. 0 prompt texts contain a tracked name. 0/624 answers were analyzed without target-brand identity. Automatic checks are not independent certification.
What counts
Model API answers with web search enabled, not consumer apps. Market supplied in prompts. A mention is not a recommendation. Counted once per answer. A recommendation is a derived label based on the answer wording. Unknown analyses are not treated as absences.
What may vary
Consumer apps and different prompts may produce different results. Repeat spread describes these prompts, not a population confidence interval. Different models or analysis versions can prevent like-for-like comparisons.
What to verify
Machine-derived labels can be wrong. Quotes are checked against original text; this validates the quotation, not its interpretation. Domain-shaped Gemini source titles are labelled as unverified destinations.
Analysis version: report-v2.7-explicit-identities. Preserve your raw-answer export alongside this report. Selected supporting answers follow; the full dataset remains in your private report.
Not affiliated with OpenAI, Anthropic, Google or Perplexity.