GeoAura

Trust & method

Corrections & Methodology

In citation work, a publisher that corrects in public is more useful than one that has never been wrong — because the second claim is not verifiable. This page lists every correction we have made since September 2026, with the original wording left visible, the reason it was wrong, and the primary source used to replace it. It also publishes the standard we apply to quotations, and the third-party figures by which we are measured.

Our correction rule

We withdraw, we do not quietly overwrite. When a claim is wrong, the correction is left on the page that carried it, with the date, so a reader who quoted the old version can see what changed and when.

We do not keep numbers we cannot source. If a precise figure cannot be traced to a named, checkable source, it is deleted rather than rephrased. Losing a number costs less than keeping an unsourced one.

Single-tactic research is never combined. If a study measured tactic A and tactic B separately, we do not add, multiply or extrapolate them. The Princeton GEO paper measured nine tactics individually; only one pair (fluency × statistics) was tested together, at +5.5%.

How we verify quotations

Every quotation in every article is classified into one of four states. The ledger is regenerated from the source files on every build, so a quotation that loses its provenance is caught rather than shipped.

StateRuleShips?
VerbatimThe exact sentence appears in a named, linkable source — arXiv id, DOI or page number quoted alongside.Yes
ReproducibleQuotation marks retained because any reader can open the linked PDF or dataset and find the wording themselves.Yes
ParaphraseOur own analysis, or an explicitly labelled annotation of a source. No quotation marks, no implication that the source said it.Yes
UnverifiedQuotation marks plus a named source that does not contain the sentence. Blocked in review; treated as a defect, not a style choice.No

Correction log

Seven entries, newest first. Each one names the page, the wording that was wrong, why it was wrong, and the source used to replace it.

2026-10-28Homepage meta description

Original"…expert quotations boost AI visibility by 41%, statistics by 33%."

What was wrong Both figures are real and separately measured, but placed in one sentence they read as a combined lift — the exact impression we retracted on 26 October. The field is machine-read: a third-party index snapshotted the sentence verbatim on 21 October.

Corrected to Rewritten so the sentence states that Princeton measured each tactic separately and never tested them as a combined lift, with a pointer to this page.

Basis: Aggarwal et al., arXiv:2311.09735; external index snapshot read_on 2026-10-21

2026-10-279 Proven GEO Strategies · How to Get Cited by AI Search

Original"Don't try to rank. Try to be cited." — attributed to the Princeton GEO research team

What was wrong The sentence does not exist in the paper at any level. It was our own editorial slogan carrying a research attribution.

Corrected to Replaced with the paper abstract's verbatim sentence, marked verified, with an arXiv link. The slogan is retained only where it is labelled as a GeoAura position, not a finding.

Basis: Aggarwal et al., arXiv:2311.09735 (abstract)

2026-10-27Content Structure for AI Search

Original"Numbered lists are the most extractable format…" — attributed to a Princeton GEO-bench extraction analysis

What was wrong No such conclusion exists in the paper; extraction-format preference was never tested.

Corrected to Downgraded to a labelled editorial reading of the paper ("GeoAura reading — editorial interpretation, not a verbatim extract").

Basis: Aggarwal et al., arXiv:2311.09735 (full text)

2026-10-27AI Search Optimization · Does Domain Authority Matter

OriginalQuotation attributed to "Prabhat Agarwal, lead author of the Princeton GEO study"

What was wrong Two errors: the sentence is not in the paper, and the author's name is wrong (Pranjal Aggarwal). A wrong byline is worse than a wrong number, because it is the field a fact-checker reads first.

Corrected to Quote downgraded to a summary with no quotation marks; the byline corrected site-wide. "Dugan, L." remains a live search-and-destroy term in our build checks.

Basis: Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K. & Deshpande, A., arXiv:2311.09735

2026-10-2739 instances across the site (incl. Princeton benchmark · GEO audit checklist)

Original"GEO optimization lifts a rank-5 page +115% and cuts a rank-1 page −30%"

What was wrong The numbers are real but the scope was overstated. They belong to one method (Cite Sources), not to GEO as a whole, and −30.3% is an average across rank-1 pages.

Corrected to Rewritten as "+115.1% … under the Cite Sources method" and "rank-1 pages averaged −30.3%", with section and table references.

Basis: Aggarwal et al., arXiv:2311.09735, §5.2 and Table 2

2026-10-27How to Measure GEO Visibility

Original"Only 16% of Fortune 500 brands track AI visibility"

What was wrong The precise figure could not be located in the cited AirOps report. An exact number with no traceable source is not a statistic.

Corrected to Deleted. The page now states only what the report actually supports, without a percentage.

Basis: AirOps public report (re-checked 27 October 2026)

2026-10-269 Proven GEO Strategies · Statistics for AI Citations · 3 further pages + meta

Original"Statistics + citations compound to +61% visibility"

What was wrong An extrapolation, not a measurement. Princeton tested statistics and citations separately and never as a pair; the only combination the paper scored is fluency × statistics at +5.5%. The claim was first flagged by an external audit (GeoParrot).

Corrected to Withdrawn on every page and in metadata, with the retraction left visible rather than silently deleted, plus an explicit rule: no page may multiply, add or extrapolate single-tactic figures.

Basis: Aggarwal et al., arXiv:2311.09735 (Table 1); external audit by GeoParrot, 25 October 2026

External measurement: what third parties publish about this site

These are read from an independent index, not from our own dashboard. We publish them because a self-reported visibility number is worth nothing, and we do not rewrite figures we did not produce — including the ones that look unflattering.

MeasurementReadingAs of
Citation rate (all engines)0.04% — 12 citations across 31,659 observed runs2026-10-26 release
Most recent citation2026-10-262026-10-26 release
Best-performing query shape4.23% — rank #32 of 684 domains in the "is it worth it" shape2026-10-26 release
Average citation position3.1 (best: 1)2026-10-26 release

Source: Machine Relations Index (mri_score_v2.1+2026-10-26+ff67a630fa8e, artifact sha256 b37ddb40…57e1, window 2026-05-10 → 2026-10-26, 33,944 domains, 222,880 source events). Public listing: machinerelations.ai/index/domains/geoaura.world (Zenodo DOI 10.5281/zenodo.23652423). Figures are reproduced as published; where the index's per-engine and monthly tables disagree, we say so rather than reconcile them.

Report an error

If a page on this site misstates a number, misattributes a quotation or overstates what a source found, send the page URL and the source we should have used to our contact page. Corrections are published here with the date, and the original wording stays visible.

This page is maintained by hand and dated on every change. Last updated 28 October 2026.