Lighthouse AdvisorySLED AI Adoption Intelligence
← Back to results

From the Google edition of October 29, 2026

Independent reportingMixedNew this fortnight

Technical.ly small test: Gemini 3.5 Flash-Lite answered three Pittsburgh election questions accurately and declined to endorse, while ChatGPT made one slip; outlet still advises primary sources

Technical.ly · Local government (county election information) · Pittsburgh / Allegheny County, Pennsylvania

Publisher
Technical.ly
Original publication
2026-10-21
Source retrieved
2026-10-30
Event date
2026-10-19
Read original source

What happened

On Oct. 19 Technical.ly asked ChatGPT and Gemini (3.5 Flash-Lite) three conversational Pittsburgh election prompts in fresh, signed-out sessions. Both declined to tell users how to vote. Gemini was accurate on all three, more concise, mostly pointed users to existing voter guides and warned that election information changes quickly; ChatGPT referenced a ballot question already decided in the primary. The outlet concludes chatbots remain a shaky bet and primary sources are safer, citing ISD's finding that 29% of responses across six chatbots were incomplete, inaccurate or outdated.

Why it matters

Complements the statewide tracker with a local-election view: Gemini tended to defer to existing guides, reinforcing the role of county election offices and local voter guides as the sources assistants surface.

Evidence and measured results

Three prompts per model, one date, one locality; qualitative comparison. Not statistically meaningful.

Limitations and uncertainty

Very small informal test; different Gemini model (3.5 Flash-Lite) from the tracker (2.5 Flash); results may not generalize.

Put this evidence to work

Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-30; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.

Sales

Role takeaway
Customer problem
County election offices want voters to reach accurate local information, increasingly via AI assistants.
Stakeholders
  • County elections director
  • County communications office
Discovery questions
  • Are local guides and sample ballots easy for assistants to find?
  • Has the office tested assistant answers about local races?
Value hypothesis
Clear official pages may increase the chance assistants route voters correctly; not measured.
Potential engagement
Small local AI-answer spot check and content tune-up.
Unsupported claims
  • General Gemini election accuracy
  • Superiority over other assistants

Pre-sales engineering

Role takeaway
Fit
Relevant to county election web content owners.
Architecture and integration
None; public web content only.
Prerequisites
  • Current official guide and sample-ballot pages
Constraints
  • Tiny sample
  • Model variation over time
Security
Monitor for outdated pages that assistants might surface.
Proof of value
  • Repeat local prompts across assistants and dates

Delivery

Role takeaway
Work
Local assistant spot checks and content updates.
Dependencies
  • Elections staff verification
Ownership
County communications office.
Skills and adoption
Brief staff on testing assistants.
Governance checkpoints
  • Elections director review of findings
Acceptance criteria
  • Proposed: documented spot-check results before Election Day
Risks
  • Generalizing from three prompts

Implementation considerations

Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.

Architecture and integration

What must connect, and where does the AI sit in the workflow?

Assistants route users to existing voter guides; local offices benefit from well-indexed official sample-ballot and guide pages.

Governance

Who approves, reviews and stays accountable for outcomes?

Neutrality behavior (declining endorsements) observed; local officials should still direct voters to official sources.

Security and privacy

What data, permissions and controls need testing?

No security testing; signed-out sessions limit personalization effects.

Accessibility and workforce

Who is affected, and what skills or accommodations follow?

Concise answers may help casual users but language access was not tested.

Procurement

What should contracts, pricing and exit terms secure?

None directly; informs public-communication strategy rather than purchasing.

Operating model

Which teams own the service once it runs?

Supports a communications approach in which AI assistants are treated as pointers to official sources, not authorities.

Publication history

  1. 2026-10-29Google · Issue 018 resources
Read preserved resource versions (JSON)

Stable resource ID: technically-pittsburgh-election-test-gemini-chatgpt-20261021