SEO Recap covering September 7, 2026: Harvard's AI job automation study and conflicting AI answers

Built by Stephanie Chung·4 min read·3 stories

Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)

Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)

Wrong or outdated AI answers about a brand usually stem from too many conflicting versions of the truth (old PDFs, stale bios, retired product docs), not from a lack of published content. AI engines retrieve those sources and construct one settled-looking answer.
The prompt supplies the vocabulary that drives retrieval, so a question like "Who is the CEO?" surfaces old pages using that exact term while current pages using "SVP and GM" or "brand president" never enter the retrieval set.
Fix it with bridge content that connects obsolete language to present reality, for example "Following the acquisition, the company no longer has a standalone CEO. Jane Smith now leads it as SVP and general manager."
Run a brand claim audit, not just a visibility audit, documenting for each key claim:
- The likely prompt and any outdated assumption it contains
- The old terminology and its current equivalent
- The approved current fact and its canonical source
- Every owned page, PDF, feed, or profile carrying an older version
- The action needed (update, annotate, consolidate, redirect, retire, or bridge)
Measure answer accuracy, not just presence. A brand mention is not a win if the answer names a former executive, an old price, or a discontinued feature. Inspect citations and reproduce the searches before assuming a hallucination.

James Riley's October 2025 study scored 940 occupations on how morally objectionable full automation would be (1-7 scale) across 2,357 respondents. Search marketing strategists scored 2.31, with only file clerks lower. Clergy scored 5.91 and childcare workers 5.86.
Public support for full automation ran ~30% of tested jobs at current AI capability, and nearly doubled to 58% when respondents imagined a cheaper AI that outperforms humans. Only ~12% of jobs drew strong resistance regardless of performance.
Elisabeth Paulson's conjoint experiment (9,000 participants) found people leaned human by 4.3 points on loan approvals and 7.6 points on pretrial release, but preference tracked belief about competence. Among those who thought algorithms were better, 54-56% picked the algorithm.
A Procter & Gamble study of 791 developers found top-10% quality ideas were 3 times more likely from AI-assisted teams than unassisted individuals, evidence the competence gap protecting SEO work is closing.
Practical takeaways for AI-touched content:
- Attach a real, verifiable human byline with credentials, not a generic "Editorial Team" label
- Publish your measurable performance record inside the piece as a competence signal
- Reserve full automation for repetitive "no-joy" tasks like link audits and meta drafts, keeping a named human on anything touching reader trust or client money

John Mueller said programmatic SEO "often leads to a site that's either spam, borderline spam, or low quality," writing that it is easy to spin up many pages but hard to provide real value to users.
He warned that Google's systems can "lose faith" in a site's value based on those old pages, and recovery "tends to take time and significant effort to show the value."
Mueller compared the recovery timeline to resolving spam and core update issues, meaning many months of work to rebuild trust and ranking.

Published on September 9, 2026