Can AI citations fall while brand rankings stay stable?

Research background

The public Summer 2026 Profound Index summary examines regional visibility, model changes and brand movement. Profound describes the underlying Index as powered by 1.9+ billion real user conversations across 50+ industries.

That is the stated scope of the Index, not a disclosed sample size for every finding in this edition. This analysis focuses on citation resilience and market-specific measurement rather than reproducing the gated report.

How the research was done

This analysis uses only numerical findings visible on Profound’s public Summer 2026 report page. Profound reports a before-and-after comparison around a ChatGPT update and cross-market industry comparisons.

The public page does not disclose all prompt selections, sampling windows or uncertainty estimates. The chart reproduces the reported decline magnitude without estimating missing baselines.

Report

Profound Index Report: Summer 2026

Research period

Summer 2026; model-update findings concern June

AI platforms

ChatGPT

Original author

Profound

Key research statistics

The numbers in charts

Each chart uses the values published in the original report.

What the research actually reveals

What changed after the model update? Profound says unique cited domains declined in 56 of 58 industries, alongside the 10.4% fall in citations per answer.

Yet nine of ten leading brands in the median industry retained their position. Together, these observations show why a stable leaderboard can conceal a change in the supporting evidence layer. Rank and source diversity answer different questions.

The public summary also reports that in 24 of 30 multi-region industries, a regional market had a leading brand that did not appear among the reference market’s leading options.

A central visibility score can therefore obscure important variation. The finding does not establish the same pattern for every software category, but it argues against treating one market’s prompt panel as a universal benchmark.

In markets with a new regional leader, 46–59% of citations went to market-specific domains. This is a reported range, not a pooled average.

It supports examining which sources are relevant in each market rather than merely translating owned pages. The chart deliberately shows the separate citation-volume decline instead of turning the range into a fabricated midpoint.

Implications for Series A to Series C SaaS companies

For Series A–C B2B SaaS companies, a stable brand position is not sufficient evidence that an AEO programme is healthy.

If the publications supporting the brand disappear, the remaining visibility may depend on a narrower evidence base. That is a resilience concern, not proof of an impending ranking collapse.

An expanding SaaS business should assess visibility by market, language and buyer situation. It should also distinguish named-product mentions, linked citations and actual referral outcomes.

A change in one should not be presented to leadership as an equivalent change in another.

Actionable recommendations

1. Maintain separate views for brand inclusion, rank, citation count and cited domains. Save the underlying answers so that changes can be traced to specific sources rather than inferred from a dashboard average. Annotate model updates and compare like-for-like prompts before drawing conclusions.
2. Audit whether important product claims depend on a narrow set of external publications. Improve the accuracy and accessibility of owned evidence, and pursue relevant independent coverage without assuming that any placement guarantees a recommendation. Review market-specific sources before expanding a translated content programme. AEO Operators can help turn these checks into an ongoing visibility and source-resilience programme.

Original sources and further reading

• Profound Index Report: Summer 2026

  Profound, Profound, 19 August 2026. Section: Public report page. Vendor research.

  The publisher’s public report page supplies every numerical finding used in this analysis.