Parse
IndexPricing
Sign inCheck your brand
  1. Research
Research
Does AI criticize its top recommendation?

Does AI criticize its top recommendation?

Published Jul 30, 2026Updated Oct 12, 2026

On this page
  • The finding
  • How we measured
  • Sources
  • More like this
  • The finding
  • How we measured
  • Sources
  • More like this

Dimitry ApollonskyFounder, Parse

DataWhat reviewed means

A first-place recommendation can still include a drawback. We wanted to see how often AI criticizes the brand it ranks first, rather than treating the top position as an unqualified endorsement.

We examined reviewed statements about top-ranked brands in ChatGPT Search and Google AI Mode answers from May 24 to July 15, 2026. We counted specific criticism attached to those recommendations, keeping recommendation position and descriptive tone as separate measures.

AI sometimes names a drawback even for its first choice. Specific criticism appeared in 3.6% of the top-ranked recommendations we analyzed from ChatGPT Search and Google AI Mode.

AI criticized 3.6% of its top recommendations

A specific negative claim appeared with the top-ranked brand in 5,460 of 150,193 analyzed recommendations, or 3.6%. The other 144,733 top recommendations, or 96.4%, had no specific negative claim about that brand. A negative claim is a concrete drawback, limitation, risk, or unfavorable comparison stated in the answer.

First place is not always an unqualified endorsement. Track recommendation rank and the claims around the brand separately so a high rank does not hide a repeated limitation.

Takeaway

Review the language around a top-ranked brand before treating first place as an unqualified endorsement.

ChatGPT Search criticized top picks 5.5 times as often

ChatGPT Search included a negative claim in 4,461 of 67,472 top recommendations, or 6.6%. Google AI Mode did so in 999 of 82,721, or 1.2%. ChatGPT Search's rate was 5.5 times as high, a gap of 5.4 percentage points.

A combined criticism rate hides a large engine difference. Audit the same priority prompts on both engines and keep separate baselines for each.

Criticism rate by engine

  • ChatGPT Search6.6%4,461 of 67,472
  • Google AI Mode1.2%999 of 82,721
  • 0%2%4%6%8%

Takeaway

Use an engine-specific criticism baseline instead of one blended rate.

Most criticism did not make the whole recommendation negative

Of the 5,460 criticized top recommendations, 4,912, or 90.0%, did not carry negative overall sentiment. Across all 150,193 top recommendations, 560, or 0.37%, had negative overall sentiment; 2,792, or 1.9%, were reluctant; and 148, or 0.10%, explicitly said the brand was not recommended for the need.

A sentiment label alone misses most specific drawbacks. Keep negative claims, overall sentiment, reluctance, and explicit rejection as separate fields because they answer different questions.

included a specific negative claim
3.6%

5,460 of 150,193

had negative overall sentiment
0.37%

560 of 150,193

were reluctant recommendations
1.9%

2,792 of 150,193

explicitly said the top brand was not recommended
0.10%

148 of 150,193

Criticism became more common lower in the recommendation list

Negative claims appeared with 5,460 of 150,193 first-place recommendations, or 3.6%; 17,989 of 259,382 brands ranked second or third, or 6.9%; and 32,388 of 377,610 brands ranked fourth or lower, or 8.6%. The fourth-or-lower rate was 2.4 times the first-place rate.

Rank still carries information about how the answer frames a brand. Compare criticism among similar recommendation positions instead of treating every named brand as equivalent.

Criticism rate by recommendation position

  • First3.6%5,460 of 150,193
  • Second or third6.9%17,989 of 259,382
  • Fourth or lower8.6%32,388 of 377,610
  • 0%5%10%

Takeaway

Benchmark criticism among brands at similar recommendation positions.

Software's top picks drew criticism 2.7 times as often as consumer goods

Among industries with at least 1,000 analyzed top recommendations, Software recorded 414 criticized top picks in 8,448, or 4.9%. Consumer Goods recorded 41 in 2,219, or 1.8%. The gap was 3.1 percentage points, and the Software rate was 2.7 times as high.

Use an industry baseline before treating a brand's rate as unusual. This spread identifies where to review answer language and does not show that industry caused the criticism.

Selected industry criticism rates

  • Software4.9%414 of 8,448
  • Data and Analytics4.5%138 of 3,076
  • Information Technology4.4%310 of 7,103
  • Financial Services4.1%377 of 9,140
  • Consumer Electronics2.4%33 of 1,377
  • Travel and Tourism2.1%31 of 1,504
  • Sports2.0%35 of 1,793
  • Consumer Goods1.8%41 of 2,219
  • 0%2%4%6%

Datadog had the most criticized top recommendations

Among brands with at least 30 criticized top recommendations, Datadog led by count with 213 of 1,605, or 13.3%. Atlassian followed with 139 of 900, or 15.4%. Playwright had the highest rate in the displayed set at 71 of 398, or 17.8%.

These are audit starting points, not brand-quality scores. Prompt mix differs by brand, so compare a brand with its own prompts and recurring limitations before comparing rates across brands.

Criticism rates for named brands

Criticized (%) · 10 results

  • Playwright17.84%
  • Atlassian15.44%
  • Upwork14.83%
  • Datadog13.27%
  • Linear9.71%
  • Profound8.8%
  • Stripe8.25%
  • Grafana7.85%
  • Bright Data6.64%
  • LinkedIn5.59%
  • 0%5%10%15%20%
Explore the data (10 rows)
Datadog2131,60513.271111
Atlassian13990015.4444135
LinkedIn981,7545.5872167
Linear889069.71343
Playwright7139817.839230
Grafana577267.851263
Upwork5134414.825666
Profound505688.802841
Bright Data497386.639637
Stripe475708.2456118

Takeaway

Start with repeated brand-specific limitations, then read the underlying answer context before changing messaging.

The result held after narrower eligibility checks

The main result covered 150,193 of 215,354 answers with one unambiguous top-ranked brand. We excluded 65,161 top recommendations without a reviewed statement about that brand and one answer with tied top brands. The rate was 3.6% from June 1 through July 15 and 3.6% when each answer had one topic type, compared with 3.6% in the main result.

The sensitivity checks do not change which recommendations the study could include. The result describes the prompts analyzed here and does not measure recommendation accuracy, brand quality, buyer opinion, or causation.

main result
3.6%

5,460 of 150,193

June 1 through July 15
3.6%

5,459 of 150,170

one topic type per answer
3.6%

5,425 of 149,786

What marketers should do

The top-ranked brand carried a specific negative claim in 3.6% of analyzed recommendations. ChatGPT Search's rate was 5.5 times Google AI Mode's, and criticism was more common for brands ranked lower.

Track rank, specific drawbacks, overall sentiment, reluctance, and explicit rejection separately. Audit the same priority prompts on both engines. Group repeated limitations by brand and topic, then inspect the answer and cited sources before changing positioning or content.

  • Being cited is not enough: how to monitor sentiment in AI answers

How we measured

We analyzed 162,788 reviewed statements about 150,193 top-ranked brand recommendations, covering 22,546 brands and 15,858 organic prompts on ChatGPT Search and Google AI Mode from May 24 through July 15, 2026.

higher criticism rate on ChatGPT Search
5.5×

6.6% versus 1.2%

of criticized top picks were not negative overall
90.0%

4,912 of 5,460

of brands ranked fourth or lower included criticism
8.6%

32,388 of 377,610

Get the data

Dataset CSVThe metrics behind every figure in this report.

Sources

  1. Semrush: What is AI sentiment analysis? A marketer's guide (opens in a new tab) · accessed July 30, 2026
  2. Semrush: AI Visibility Brand Performance Reports (opens in a new tab) · accessed July 30, 2026
  3. Ahrefs: How to monitor and win brand mentions in AI answers (opens in a new tab) · accessed July 30, 2026
  4. The Language Blind Spot: Brand reputation across twelve European languages (opens in a new tab) · accessed July 30, 2026

More like this

How often does AI recommend a brand with reservations?
AI attached a reservation to a ranked brand in about one in ten appearances.
Does AI keep criticizing the same brand?
Criticism usually did not repeat in the next answer about the same brand.
Does AI praise and criticize the same brand?
AI sometimes praised and criticized the same brand within one answer.
How often does AI recommend against a brand?
AI rarely explicitly advised against a brand for the buyer's stated need in the statements we reviewed.
Does AI make brands sound better than cited sources?
AI was more likely to make brands sound better than worse relative to their cited sources.
What words does AI use to describe brands?
Most brand descriptions were positive, and much of the praise was shared across brands.
Does AI cite sources when it criticizes a brand?
Most criticism of ranked brands came without a citation attached to the criticized brand.
Does AI praise brands even when it ranks them lower?
Praise was common even for brands ranked fourth or lower, so positive wording alone did not signal first choice.
Does AI recommend the same brand when you ask again?
Asking the same engine the same question again kept the first recommendation in only about six in ten comparisons.
Do ChatGPT and Google criticize the same brands?
When one engine criticized a brand both ranked, the other usually did not share that criticism.
Where does AI make a clear recommendation?
How often AI expressed reservations varied by category. Reservations appeared in 1.6% of reviewed ranked-brand appearances in warehouse robotics, compared with 22.6% in Ethereum DeFi tokens.
What reasons does AI give for its first recommendation?
AI cited reputation more often for first choices than lower-ranked brands. Price was the first choice's most common reservation.
Which brands does AI tell you to avoid?
AI named brands to avoid much more often when explicitly asked. Unprompted warnings were less common.

See how AI ranks and describes your brand.

Parse

Index

  • Brands
  • Markets
  • Topics
  • How we measure

For agents

  • Docs
  • Connect
  • Workflows
  • llms.txt

Parse

  • Pricing
  • Research
  • Blog
  • Compare

Company

  • About
  • Contact
  • Corrections

© 2026 Parse, Inc. All rights reserved.

Privacy PolicyTerms of Service