Parse
IndexPricing
Sign in
  1. Research
Research
Tables vs lists vs prose: how AI answer formatting doubles the brands named

Tables vs lists vs prose: how AI answer formatting doubles the brands named

Published Nov 20, 2026

On this page
  • Layout split
  • Brands per layout
  • Same-prompt control
  • Recommendation dilution
  • Treatment shift
  • Incidental mentions & tone
  • Citations & length
  • By industry
  • What to do
  • Study scope
  • How we measured
  • More like this
  • Layout split
  • Brands per layout
  • Same-prompt control
  • Recommendation dilution
  • Treatment shift
  • Incidental mentions & tone
  • Citations & length
  • By industry
  • What to do
  • Study scope
  • How we measured
  • More like this

Dimitry ApollonskyFounder, Parse

DataWhat reviewed means

Does the layout an AI engine chooses, such as a Markdown table, a numbered list, bullet points, or plain prose, change how many brands appear and whether those mentions count as recommendations or side-by-side comparisons?

We classified 424,604 buyer answers on ChatGPT Search and Google AI Mode from October 15 through November 19, 2026 into five Markdown layouts, verified layout effects inside 26,162 fixed-prompt pairs, and tracked 915,973 brand judgments scored under one uniform Parse evaluator version across October 15 through October 26, 2026.

Across 424,604 buyer answers, ChatGPT Search used Markdown tables 4.8x more often than Google AI Mode, packing 81% more brands into table-plus-list layouts while shifting mentions from direct recommendations to comparisons.

ChatGPT formats 76.5% of buyer answers with tables; Google defaults to bullets (62.2%)

The two search engines choose fundamentally different answer structures for the same market of buyer questions. Across 217,543 ChatGPT Search buyer answers, 76.5% contained a Markdown table: 39.0% rendered a table only (84,850 answers) and 37.5% paired a table with a numbered list (81,640 answers). Bullets-only answers made up 14.3% (31,001), numbered-only 6.0% (13,148), and prose-only 3.2% (6,904).

Google AI Mode flips that template completely. Across 207,061 buyer answers, 62.2% used bullets only (128,859) and 20.2% used a numbered list only (41,880). Only 16.1% included any Markdown table (12.4% table only and 3.7% table plus numbered list), making ChatGPT Search 4.8x as likely to render a table. Prose-only answers were rare on both engines at 3.2% on ChatGPT and 1.4% on Google.

ChatGPT buyer answers with a Markdown table
76.5%

4.8x Google AI Mode's 16.1% table rate

Google uses bullets only in 62.2% of buyer answers

ChatGPT brands named: table + list vs prose
8.03 vs 4.44

81% more brands in table + numbered layouts

Google runs 7.64 brands vs 4.09 in prose

Same-prompt lift on ChatGPT (table + list vs bullets)
+24.6%

6.07 vs 4.87 brands on 3,106 identical prompts

+9.3% on Google (7.76 vs 7.10 across 4,006 prompts)

Google comparison share surge inside tables
4.9x

12.3% compared in table-only vs 2.5% in bullets-only

Recommended share drops from 70.3% to 51.2% in table + list

Share of buyer answers by Markdown layout and engine

  • ChatGPT Search
  • Google AI Mode
  • Table only
  • ChatGPT Search39.0%84,850
  • Google AI Mode12.4%25,678
  • Table + numbered list
  • ChatGPT Search37.5%81,640
  • Google AI Mode3.7%7,730
  • Bullets only
  • ChatGPT Search14.3%31,001
  • Google AI Mode62.2%128,859
  • Numbered list only
  • ChatGPT Search6.0%13,148
  • Google AI Mode20.2%41,880
  • Prose only
  • ChatGPT Search3.2%6,904
  • Google AI Mode1.4%2,914
  • 0%20%40%60%80%
424,604 buyer answers (217,543 ChatGPT Search, 207,061 Google AI Mode), October 15 through November 19, 2026.

Takeaway

ChatGPT Search runs as a comparison grid by default (76.5% tables), while Google AI Mode runs as a bulleted shortlist (82.4% bullets or numbered lists without tables).

Table-plus-list answers name 8.03 brands per answer, nearly double prose-only

Answer layout strongly governs how wide a brand shortlist becomes. On ChatGPT Search, answers pairing a table with a numbered list averaged 8.03 brands per answer, 81% more than prose-only answers (4.44 brands) and 73% more than bullets-only answers (4.64 brands). Table-only ChatGPT answers averaged 6.74 brands.

Google AI Mode shows the same step-up even though tables appear far less often there. Google table-plus-numbered answers averaged 7.64 brands and table-only answers averaged 7.50 brands, compared to 6.49 in bullets-only, 6.95 in numbered-only, and 4.09 in prose-only answers.

Answer footprint across all 5 layouts on both engines

Share of answers, average brands named, average citation links, and average answer length in characters for each engine and Markdown layout.

Avg brands · 10 results

  • ChatGPT Search · Table + numbered list8.03
  • Google AI Mode · Table + numbered list7.64
  • Google AI Mode · Table only7.5
  • Google AI Mode · Numbered list only6.95
  • ChatGPT Search · Table only6.74
  • Google AI Mode · Bullets only6.49
  • ChatGPT Search · Bullets only4.64
  • ChatGPT Search · Numbered list only4.61
  • ChatGPT Search · Prose only4.44
  • Google AI Mode · Prose only4.09
  • 0510
Explore the data (10 rows)
ChatGPT SearchTable + numbered list81,64037.58.0312.7210,224
Google AI ModeTable + numbered list7,7303.77.642.54,057
Google AI ModeTable only25,67812.47.534,075
Google AI ModeNumbered list only41,88020.26.952.313,521
ChatGPT SearchTable only84,850396.7412.018,174
Google AI ModeBullets only128,85962.26.492.362,928
ChatGPT SearchBullets only31,00114.34.647.915,329
ChatGPT SearchNumbered list only13,14864.617.335,766
ChatGPT SearchProse only6,9043.24.448.235,093
Google AI ModeProse only2,9141.44.090.782,806
424,604 buyer answers, October 15 through November 19, 2026. Filter by engine or switch to Chart view.

Takeaway

Structured grids hold space for roughly 7.5 to 8 brands per answer, while prose collapses the shortlist to roughly 4 brands.

Even on the exact same buyer prompt, switching to a table + list adds up to 24.6% more brands

Aggregate averages could simply reflect different questions triggering different layouts. To isolate formatting from topic mix, we tested prompts observed on different days with both a table layout and a non-table layout on the exact same question.

Across 10,430 ChatGPT Search prompts observed in both formats, answers with any table named 5.63 brands vs 4.95 without a table (+13.7%). Narrowing to the 3,106 ChatGPT prompts seen as both table + numbered list and bullets only, table + numbered answers named 6.07 brands vs 4.87 in bullets (+24.6%). On Google AI Mode, the same fixed-prompt check delivered +5.1% more brands across 15,732 table vs non-table prompts (7.40 vs 7.04) and +9.3% across 4,006 table + numbered vs bullets prompts (7.76 vs 7.10).

Within-question fixed-prompt brand count lift

Brand lift (%) · 4 results

  • ChatGPT Search · Table + numbered vs Bullets only24.6%
  • ChatGPT Search · Any table vs Non-table13.7%
  • Google AI Mode · Table + numbered vs Bullets only9.3%
  • Google AI Mode · Any table vs Non-table5.1%
  • 0%10%20%30%
Explore the data (4 rows)
ChatGPT SearchTable + numbered vs Bullets only3,1066.074.8724.6
ChatGPT SearchAny table vs Non-table10,4305.634.9513.7
Google AI ModeTable + numbered vs Bullets only4,0067.767.19.3
Google AI ModeAny table vs Non-table15,7327.47.045.1
Same buyer prompt observed under both layouts during October 15 through November 19, 2026.

Takeaway

Format itself expands the shortlist: holding the buyer prompt identical, rendering a table with a numbered list opens up to 24.6% more brand slots.

Tables dilute recommendation status from 70.3% down to 51.2% on Google AI Mode

Naming more brands inside a table comes with a clear tradeoff: fewer of those mentions are direct recommendations. Across 915,973 brand judgments scored under a single uniform Parse evaluator version from October 15 through October 26, 2026, Google AI Mode's recommended share dropped from 70.3% in bullets-only answers and 69.3% in prose-only answers down to 55.3% in table-only answers and 51.2% in table-plus-numbered answers.

ChatGPT Search kept recommendation share higher inside table-only answers (68.9%, matching 68.4% in prose-only and 68.1% in bullets-only), before dipping to 62.1% in table-plus-numbered answers and 62.7% in numbered-only lists.

Share of brand mentions classified as recommended, by layout (Oct 15–26)

  • ChatGPT Search
  • Google AI Mode
  • Bullets only
  • ChatGPT Search68.1%
  • Google AI Mode70.3%
  • Prose only
  • ChatGPT Search68.4%
  • Google AI Mode69.3%
  • Table only
  • ChatGPT Search68.9%
  • Google AI Mode55.3%
  • Numbered list only
  • ChatGPT Search62.7%
  • Google AI Mode66.4%
  • Table + numbered list
  • ChatGPT Search62.1%
  • Google AI Mode51.2%
  • 0%20%40%60%80%
915,973 brand judgments under one uniform Parse evaluator version (October 15 through October 26, 2026).

Takeaway

Getting named in a Google AI Mode table carries a lower chance of an outright recommendation (51.2% to 55.3%) than appearing in a standard bulleted shortlist (70.3%).

Head-to-head comparisons jump 4.9x in Google tables and 2.0x in ChatGPT tables

As direct recommendation share falls inside tables, head-to-head comparisons and incidental mentions fill the extra rows across the October 15 through October 26, 2026 uniform-evaluator window. On Google AI Mode, compared brand mentions jumped 4.9x from 2.5% in bullets-only (and 0.6% in prose-only) up to 12.3% in table-only answers and 11.5% in table-plus-numbered answers.

ChatGPT Search showed a 2.0x jump in compared brand share, rising from 3.3% in prose-only and 3.8% in bullets-only to 5.5% in table-only and 6.7% in table-plus-numbered answers. ChatGPT lists without tables carried the highest alternative-framing rates, reaching 11.1% in numbered-only and 10.5% in bullets-only vs 2.3% to 3.1% across all Google layouts.

Brand treatment mix across all 10 engine-layout cells (915,973 judgments, Oct 15–26)

Share of brand mentions judged as recommended, compared, alternative, incidental, or price/spec only across 915,973 uniform-evaluator observations (October 15 through October 26, 2026).

Recommended (%) · 10 results

  • Google AI Mode · Bullets only70.3%
  • Google AI Mode · Prose only69.3%
  • ChatGPT Search · Table only68.9%
  • ChatGPT Search · Prose only68.4%
  • ChatGPT Search · Bullets only68.1%
  • Google AI Mode · Numbered list only66.4%
  • ChatGPT Search · Numbered list only62.7%
  • ChatGPT Search · Table + numbered list62.1%
  • Google AI Mode · Table only55.3%
  • Google AI Mode · Table + numbered list51.2%
  • 0%20%40%60%80%
Explore the data (10 rows)
Google AI ModeTable only42,90955.312.32.523.45.8
Google AI ModeTable + numbered list6,86151.211.53.128.24.9
ChatGPT SearchTable + numbered list68,32062.16.77.722.20.9
ChatGPT SearchTable only271,49668.95.57.316.41.4
ChatGPT SearchNumbered list only13,91562.7411.119.81.8
ChatGPT SearchBullets only66,41368.13.810.5134.1
ChatGPT SearchProse only24,11368.43.38.713.65.2
Google AI ModeNumbered list only63,72166.42.82.925.91.2
Google AI ModeBullets only353,32770.32.52.321.33
Google AI ModeProse only4,89869.30.61.420.87.7
Uniform Parse evaluator version, October 15 through October 26, 2026. Filter by engine or compare shares in Chart view.

Takeaway

Tables turn brand mentions into side-by-side comparisons and incidental context rather than pure endorsements.

Table-plus-list answers push incidental mentions to 28.2% and soften attitude scores

Combining a table with a numbered list produces the highest incidental brand share on both engines across the October 15 through October 26, 2026 uniform-evaluator subset: 28.2% on Google AI Mode (vs 21.3% in bullets-only) and 22.2% on ChatGPT Search (vs 13.0% in bullets-only). More brands get swept into supporting cells, integration notes, and baseline examples without earning a recommendation.

Attitude scores over the same October 15 through October 26 evaluator window track the same pattern. Prose-only answers carried the warmest average attitude scores on both ChatGPT Search (2.334) and Google AI Mode (2.620), while table-plus-numbered layouts carried the lowest attitude scores on both engines (2.072 on ChatGPT and 2.311 on Google).

Incidental share in Google table + numbered answers
28.2%

Up from 21.3% in Google bullets-only (Oct 15–26 evaluator window)

ChatGPT jumps from 13.0% (bullets) to 22.2% (table + numbered)

ChatGPT mean attitude: table + numbered vs prose
2.072 vs 2.334

More table slots bring cooler, neutral descriptions (Oct 15–26)

Google shifts similarly from 2.620 (prose) to 2.311 (table + numbered)

Google price/spec mention share in prose-only answers
7.7%

Highest price/spec share across all 10 layouts (Oct 15–26)

vs 0.9% in ChatGPT table + numbered answers

Takeaway

Longer table-plus-list shortlists pull in extra brands at lower praise intensity and higher incidental mention rates.

Table layouts run longer and attach 12.72 citations per answer on ChatGPT

Tables also coincide with far more web citations and longer copy. On ChatGPT Search, table-plus-numbered answers averaged 12.72 citations and 10,224 characters, while table-only answers averaged 12.01 citations and 8,174 characters. Non-table ChatGPT formats dropped sharply to 7.33 citations in numbered-only (5,766 chars), 7.91 in bullets-only (5,329 chars), and 8.23 in prose-only (5,093 chars).

Google AI Mode attaches fewer links overall, yet its table answers still pull ahead: table-only averaged 3.00 citations (4,075 chars) and table-plus-numbered averaged 2.50 citations (4,057 chars), compared to 2.36 in bullets-only (2,928 chars) and just 0.78 citations in prose-only (2,806 chars).

Average citations per buyer answer, by layout and engine

  • ChatGPT Search
  • Google AI Mode
  • Table + numbered list
  • ChatGPT Search12.72
  • Google AI Mode2.50
  • Table only
  • ChatGPT Search12.01
  • Google AI Mode3.00
  • Prose only
  • ChatGPT Search8.23
  • Google AI Mode0.78
  • Bullets only
  • ChatGPT Search7.91
  • Google AI Mode2.36
  • Numbered list only
  • ChatGPT Search7.33
  • Google AI Mode2.31
  • 051015
424,604 buyer answers, October 15 through November 19, 2026.

Takeaway

When an engine builds a Markdown table, it pulls more supporting citations to populate the cells.

Software and Apps use tables nearly twice as often as Food and Commerce

Table formatting is not evenly distributed across buyer categories. Technical and B2B software verticals trigger structured grids at roughly double the rate of consumer product categories.

Grouped by each question's primary-market industry group, Software (56.7%), Apps (56.6%), and Data and Analytics (55.7%) used Markdown tables across more than half of combined answers, nearly double the rate of Commerce and Shopping (28.7%) and Food and Beverage (28.6%).

Share of answers containing a Markdown table, by industry

  • Software56.7%
  • Apps56.6%
  • Data and Analytics55.7%
  • Commerce and Shopping28.7%
  • Food and Beverage28.6%
  • 0%20%40%60%
Combined buyer answers across ChatGPT Search and Google AI Mode by primary-market industry group, October 15 through November 19, 2026.

Takeaway

B2B software and data teams face comparison grids by default; consumer categories still compete inside bulleted shortlists.

What to do with these numbers

Publish structured, verifiable comparison facts for ChatGPT Search and tight differentiation bullets for Google AI Mode. Because 76.5% of ChatGPT buyer answers include a Markdown table and average over 12 citations, clear feature matrices, pricing tiers, deployment limits, and integration tables give ChatGPT clean attributes to pull into comparison rows.

On Google AI Mode, where 82.4% of answers run as bulleted or numbered lists without tables, focus on crisp one-line positioning statements that earn a direct recommendation rather than getting relegated to incidental table cells. Separate plain mention counts from actual recommended status when reporting visibility to leadership.

Content formatting playbook by AI engine

1. Page structurePublish spec tables, tier limits, and comparison pages ready for grid extractionLead with concise one-sentence category fit and pros/cons bullet points
2. Brand goalEarn inclusion across 6.74 to 8.03 grid slots and protect compared columnsWin the 70.3% direct recommendation status inside bulleted shortlists
3. Reporting ruleSplit direct recommendations (62.1% in table + list) from incidental mentions (22.2%)Watch table dilution: recommended share drops from 70.3% to 51.2% in tables
Grounded in 424,604 buyer answers (October 15 through November 19, 2026) and 915,973 uniform-evaluator brand judgments (October 15 through October 26, 2026).

Takeaway

Optimize product pages with extractable comparison tables for ChatGPT Search while sharpening single-line fit claims for Google AI Mode's bulleted shortlists.
  • Plan content for AI table and list extraction
  • Audit engine layouts and brand treatments via Parse MCP

Study scope and counting rules

The main formatting dataset classifies 424,604 non-empty buyer answers across ChatGPT Search (217,543 answers) and Google AI Mode (207,061 answers) recorded from October 15 through November 19, 2026. Answers are grouped into five mutually exclusive layouts detected from Markdown syntax: table + numbered list, table only, numbered list only, bullets only, and prose only.

Within-question fixed-prompt controls restrict the comparison to exact buyer prompts observed with both layouts on different runs during the same October 15 through November 19 window: 10,430 ChatGPT Search prompts and 15,732 Google AI Mode prompts comparing any table vs non-table layout, plus 3,106 ChatGPT Search and 4,006 Google AI Mode prompts comparing table + numbered list vs bullets only.

Brand treatment shares (recommended, compared, alternative, incidental, and price/spec) and mean attitude scores come from 915,973 brand judgments scored under a single uniform Parse evaluator version (jev-1.13.0 / answer-matrix-v2), which covers the October 15 through October 26, 2026 subset of the main window. Table layout detects Markdown pipe tables; plain text aligned solely with spaces is not counted as a table. Industry table rates aggregate combined ChatGPT Search and Google AI Mode buyer answers by each question's primary-market industry group across October 15 through November 19, 2026.

How we measured

We analyzed 424,604 buyer answers (217,543 on ChatGPT Search and 207,061 on Google AI Mode) from October 15 through November 19, 2026, paired with within-question layout controls across 10,430 ChatGPT Search prompts and 15,732 Google AI Mode prompts plus 915,973 brand-treatment judgments scored under one uniform Parse evaluator version across October 15 through October 26, 2026.

ChatGPT Search buyer answers containing a Markdown table
76.5%

39.0% table only and 37.5% table + numbered list (vs 16.1% on Google AI Mode)

Brands named in ChatGPT table + numbered answers vs prose-only
8.03 vs 4.44

81% more brands per answer (7.64 vs 4.09 on Google AI Mode)

Within-question brand lift for table + numbered vs bullets on ChatGPT
+24.6%

6.07 vs 4.87 brands across 3,106 identical buyer prompts

Higher compared-brand share in Google tables vs bullets
4.9x

12.3% compared in table-only answers vs 2.5% in bullets-only

Get the data

Full study dataset (CSV)Layout distribution and within-question fixed-prompt controls (October 15 through November 19, 2026), plus uniform-evaluator brand treatment shares (October 15 through October 26, 2026) across ChatGPT Search and Google AI Mode.

More like this

How many brands does AI name in one answer?
Answers naming brands typically included five. Answers naming just one brand were uncommon.
How ChatGPT's writing style changed by model version
Stock AI wording became less common across the ChatGPT model versions we compared on the same prompts.
How often does a mentioned brand rank first?
Only about one in eight brands mentioned in an AI answer was its actual pick.
How AI picks a winner in head-to-head comparisons
AI split the wins in 50.7% of two-axis matchups and 73.3% of three-plus-axis matchups.
Local AI search: why ChatGPT and Google disagree on 83% of #1 local businesses
Across 17,121 local listings and 12,380 place-drift checks, ChatGPT Search and Google AI Mode agreed on the #1 local business 17.2% of the time.
Unasked assumptions: when prompts omit company size, AI defaults to enterprise
Across 457,174 unsegmented recommendations, Google AI Mode framed picks for enterprise 3.48x more than SMB (23.3% vs 6.7%), and all 23 industries skewed enterprise.
For agents: this page as Markdown, https://parse.gl/research/ai-answer-formatting-tables-vs-lists-brand-visibility.md. JSON, MCP · free for the last 90 days, checks from $0.01
Open in ChatGPT (opens in a new tab) Open in Claude (opens in a new tab)

This page

As Markdown: https://parse.gl/research/ai-answer-formatting-tables-vs-lists-brand-visibility.md
As CSV: https://parse.gl/research/ai-answer-formatting-tables-vs-lists-brand-visibility/dataset.csv
Reuse: Figures and tables on this page are free to republish with attribution: a link to https://parse.gl/research/ai-answer-formatting-tables-vs-lists-brand-visibility beside them. A standard followed link (dofollow) is preferred.

Read and search

Any brand's shares and rivals, with 95% ranges: GET https://agent.parse.gl/v1/tools/check_brand?subject=%3Cbrand-domain%3E
Sample answers in a market: GET https://agent.parse.gl/v1/tools/search_answers?args=%7B%22filter%22%3A%7B%22market%22%3A%5B%22%3Cmkt_id%3E%22%5D%7D%7D
Cited pages in a market: GET https://agent.parse.gl/v1/tools/search_cited_pages?args=%7B%22filter%22%3A%7B%22market%22%3A%5B%22%3Cmkt_id%3E%22%5D%7D%7D
Search the index: GET https://parse.gl/api/public/search?q=<words>

Connect

Connect an assistant: https://agent.parse.gl/mcp

An agent can save work before signup: POST https://agent.parse.gl/v1/workspaces

What it costs

No account: Public figures and up to 5 examples per list, about 120 brands a day per address.
Free account: $5 in credit, your brand's first check (20 buyer questions) and a free weekly check.
After your first top-up: Exports and deeper search results. The first top-up starts at $5.
Live checks: $0.01 per question per engine ($0.05 on Gemini and Claude).
Full public answers and pages: 25 free per day, then $0.005 each.
Public answers older than 90 days: $0.001 each the first time your workspace counts or lists one, then yours.

Prices: https://parse.gl/pricing.md

Map of the site: https://parse.gl/llms.txt

See how AI answer layouts frame your brand vs competitors

Start free in your browser.
Parse

Index

  • Brands
  • Markets
  • Topics
  • How we measure

For agents

  • Docs
  • Connect
  • Workflows
  • llms.txt

Parse

  • Use cases
  • Pricing
  • Research
  • Blog
  • Compare

Company

  • About
  • Contact
  • Corrections

© 2026 Parse, Inc. All rights reserved.

Privacy PolicyTerms of Service