Study 002

How AI translates your questions

An analysis of how AI transforms user prompts into search queries. What words does AI add? What does it remove? How much does it change?

0.16%
of queries searched exactly as typed
32.0%
completely rewritten with no words in common
24.9%
had "2025" or "2026" added for freshness
10,521
prompt-to-search query pairs analyzed
Last updated · Jan 22, 2026
[01]

Rewrite Rate

What is query translation?

When you ask ChatGPT, Perplexity, or another AI assistant a question, something happens before you get an answer: the AI searches the web. But it doesn't search your exact words.

Instead, it rewrites your prompt into a search query - adding context, injecting keywords, and reshaping your intent. This invisible layer determines which results you see and which brands get discovered.

This frozen aggregate snapshot covers 11,521 query pairs, showing how AI transforms what you ask into what it searches. Prompt, query, and brand-level rows are withheld.

Academic context
This matters more than it sounds. Researchers found that simply paraphrasing a prompt can cause up to 100% difference in which brands get recommended - a shift invisible to the user. When AI also rewrites the query before searching, the compounding effect on brand visibility is significant.

The bottom line

AI almost never searches what you actually typed.

Exact Matches
0.17%
of queries unchanged
Average Overlap
25.2%
words in common
Complete Rewrites
33.0%
share no words at all
Sample Size
11,521
query pairs analyzed
Similarity Distribution
Complete Rewrite(0-15%)
3,80733.0%
Significant Rewrite(15-30%)
3,65431.7%
Moderate Change(30-50%)
2,61822.7%
Minor Change(50-75%)
1,23810.7%
Near-exact(75-100%)
2041.8%
Average Jaccard similarity: 25.2%
11,521 query pairs analyzed
[02]

Words AI Adds

The vocabulary layer

AI doesn't just rephrase your question - it injects entirely new words you never used. These additions shape which results appear and which brands get recommended.

The most common injection? "list" - AI assumes you want a curated selection, even when you asked a simple question. Years like "2026" and superlatives like "best" and "top" follow close behind.

Why This Matters

If your content doesn't include words like "best", "top", or current years, you may not match the queries AI is actually sending to search engines - even if a user's original prompt perfectly describes your product.

Shown Injections
15,436
Aggregate Words Shown
10
Top 20 Injected Words
1
best
2,91925.3%
2
list
2,84824.7%
3
2025
2,60122.6%
4
top
1,87516.3%
5
companies
1,71014.8%
6
brands
8937.8%
7
platforms
7256.3%
8
vendors
6855.9%
9
software
6245.4%
10
providers
5564.8%
Brand-like terms withheld10 aggregate words shown.
[03]

Year Injection

2026
73 queries2.5%
2025
2,579 queries87.2%
2024
303 queries10.3%
2,956 year additions across 11,521 queries

Freshness bias

When you ask about "best CRM software", AI often searches for "best CRM software 2026" - even when you did not ask for current results.

AI assumes you want fresh information, so it automatically injects the current year (or recent years) into your queries. This freshness bias shapes which content appears in responses and creates an invisible expiration date on your content.

Why This Matters

Content without year references may be filtered out, even if it's evergreen. Regularly updating your content with current year mentions can improve your visibility in AI-powered search.

Injection Rate
25.7%
of queries affected
Dominant Year
2025
87% of injections
[04]

Brand Hallucination

Phantom competitors

When you ask "what's the best email marketing tool", AI doesn't just find answers - it inserts brand names you never mentioned. Your search for "email marketing software" becomes "Mailchimp vs Klaviyo email marketing comparison".

This happens in over 1 in 10 queries. AI's training on comparison content leads it to assume brand-specific searches - even when you asked a generic question.

Most Frequently Inserted Brands
Brand-level rows are withheld from the public snapshot. The aggregate incidence rate remains available.

Brand insertions were counted only when the original user prompt contained no brand mention.

Detected across 11,521 query transformations
[05]

Format Conversion

The listicle bias

When you ask a simple question like "good project management tools", AI doesn't search those words. Instead, it transforms your query into format-specific searches like "best project management tools list 2026".

Three format keywords dominate: "list", "best", and "top" - together appearing in over half of format-injected queries.

Why This Matters

AI's format injection creates a structural advantage for comparison content. Even if a user wants your specific product, AI searches for "best X list" - where your competitors appear alongside you, or without you entirely.

Queries Affected
73%
add format keywords
Format Types
9
distinct keywords
Format Keywords by Frequency
1
"list"
3,04526.4%
2
"best"
2,94825.6%
3
"top"
1,87716.3%
4
"review"
1701.5%
5
"vs"
1050.9%
6
"alternative"
570.5%
7
"reviews"
570.5%
8
"comparison"
540.5%
9
"alternatives"
520.5%
9 format keywords detected
[06]

Question Elimination

From questions to keywords

Users ask questions. AI searches keywords. This fundamental mismatch means your FAQ-style content may never match what AI actually queries - even when a user's question perfectly describes your product.

When you ask "What's the best way to…", AI strips the question mark, removes filler words, and constructs a keyword-dense search. The conversational tone disappears entirely.

Why This Matters

If your content is written to answer "How do I…" style questions, you may miss the keyword-based queries AI is actually sending to search engines. Consider including both conversational and keyword-rich variations.

141×
Reduction in question format
Questions stripped before search
User Prompts
35.4%
are questions
4,082 of 11,521
AI Queries
0.25%
are questions
29 of 11,521
How Questions Become Keywords
?"What's the best project management tool for remote teams?"
"project management tools remote teams 2026 comparison"
?"How do I improve my website's SEO ranking?"
"website SEO improvement tips best practices guide"
?"Which CRM software should I use for my small business?"
"CRM software small business top rated list"
Illustrative examples based on observed patterns
[07]

Audience Fabrication

Invisible targeting

You ask "what's the best CRM software" - but AI doesn't search that. Instead, it searches for "best CRM software for small business" or "best CRM for startups" - assuming an audience you never specified.

This implicit segmentation shapes which content appears. Enterprise-focused pages may be filtered out for users AI assumes are small businesses - even when that assumption is wrong.

Why This Matters

If your content targets "enterprise" but most queries get "for small business" injected, you may be invisible to AI - regardless of how relevant you actually are.

Audience Segments by Frequency
1
for beginners
6724.9%
2
for small business
6122.7%
3
for solopreneurs
2710.0%
4
for saas
248.9%
5
for agencies
238.6%
6
for enterprise
217.8%
280 total audience fabrications
[08]

Language Switching

In Prompts
6.3%
non-English
In Queries
2.4%
non-English
French
Meilleurs restaurants Paris pour dîner romantique
best romantic dinner restaurants Paris
Japanese
日本の最高のラーメン店はどこですか
best ramen shops Japan location
Spanish
Cómo preparar una auténtica paella valenciana
authentic valencian paella recipe
730 non-English prompts in sample

English bias in search

When users ask questions in French, Spanish, Japanese, or other languages, AI often translates them to English before searching - even when the user might prefer results in their native language.

This creates an invisible bias toward English-language content in AI responses, regardless of where the user is located or what language they used.

Why This Matters

If you serve non-English markets, having English versions of key content may improve your visibility in AI responses - even when users ask in their local language.

[09]

Length Delta

The expansion effect

AI doesn't just rephrase - it expands. Your concise questions become longer, more specific search queries packed with additional context, years, and qualifying terms.

The average prompt of 9.8 words becomes an 11.4-word query - a 16% expansion that fundamentally changes what gets found.

Why This Matters

Longer queries mean more specificity - and more chances for your content to mismatch. Each added word is another filter your pages must pass to appear in results.

Avg Change
+1.6
words per query
Expansion Rate
39.9%
of queries grow
Distribution
Shortened
2,49621.7%
Similar
4,42838.4%
Expanded
4,59739.9%
Word Delta Spread
-10 words
0.3%
-9 words
0.6%
-8 words
1.2%
-7 words
1.9%
-6 words
2.9%
-5 words
4.0%
-4 words
5.2%
-3 words
6.5%
-2 words
7.8%
-1 words
8.1%
0 words
9.2%
+1 words
8.9%
+2 words
8.3%
+3 words
6.6%
+4 words
6.1%
+5 words
5.2%
+6 words
4.8%
+7 words
3.6%
+8 words
3.5%
+9 words
3.1%
+10 words
2.3%
-10 words0 words+10 words
[10]

Methodology

Query Pairs
11,521

Unique prompt-query pairs after deduplication

Brands Tracked
664

Diverse sample across industries and sizes

Avg Overlap
25%

Mean word-level similarity between pairs

Rewrite Rate
33%

Queries with zero word overlap

Why this data is unique

This research is built on actual AI search behavior - the queries AI systems generate when retrieving information for real user prompts. Unlike synthetic benchmarks, every data point represents production AI behavior observed across 664 brands tracked by Trakkr.

We capture the invisible translation layer between what users ask and what AI searches. This ground-truth data reveals patterns that theoretical models miss: year injection, brand hallucination, intent escalation, and the systematic rewriting of natural language into search-optimized queries.

Source: OpenAI GPT with web_searchAnalysis: NLP tokenization + pattern matchingGenerated: Jan 22, 2026