Blog Services Free SEO Tools AI Agents Client Login Book a Call

How to Track AI Citations for Any Keyword

Zea Gatdula Zea Gatdula August 5, 2026

Key Takeaways

To track AI citations for a keyword, turn the keyword into several natural-language prompts, run those prompts across multiple AI platforms, extract every cited URL and measure which sources appear repeatedly.

The process is:

  1. Select a commercially relevant seed keyword.
  2. Define the brand, market and location.
  3. Expand the keyword into natural-language prompts.
  4. Run the prompts across relevant AI platforms.
  5. Force web retrieval when citations are required.
  6. To compare brand mentions vs backlinks, record every brand mention and cited URL alongside the page’s link data.
  7. Verify that each citation supports the answer.
  8. Group citations by domain and page type.
  9. Measure citation frequency, platform coverage and brand visibility.
  10. Repeat the research over time.

Do not make strategic decisions from one prompt or one AI response.

Use enough prompt and platform coverage to evaluate the market as a whole. A broader sample makes it easier to measure citation density by identifying recurring brands and sources within a niche.

What Does It Mean to Track AI Citations?

Tracking AI citations means recording the webpages an AI platform retrieves or references when producing an answer.

Depending on the platform, these references may appear as:

  • Linked citations
  • Source cards
  • Footnotes
  • A sources panel
  • Inline links
  • Supporting webpages
  • Search results used during retrieval

Citation tracking helps answer a more useful question than:

Is my brand appearing in ChatGPT?

It helps answer:

Where is ChatGPT, Google or Perplexity getting the information that shapes its recommendations?

This distinction matters.

A brand mention is a result.

In AEO, the sources influencing a brand mention are potential inputs you can study and influence.

Brand visibility is a lagging outcome. Cited sources function more like a leading indicator because they reveal where the AI platform obtained the information used to construct its answer.

AI Mentions and AI Citations Are Not the Same

AI mention versus AI citation.

You should track both, but they measure different things.

AI Mention

The AI response names your business, product or brand.

For example:

Recommended SEO platforms include Semrush, Ahrefs and Rankability.

Rankability has received a brand mention.

AI Citation

The response links to or identifies a webpage used to support the answer.

That source might be:

  • The brand’s own website
  • A competitor’s website
  • A directory
  • A comparison article
  • A news site
  • A YouTube page
  • A review platform
  • A forum discussion

A brand can be mentioned without its website being cited.

A website can also be cited without the business being prominently recommended.

Track the two separately.

MetricWhat it tells you
Brand mentionWhether the brand appears in the answer
Brand positionWhere the brand appears among recommendations
Citation URLWhich exact page supported the answer
Citation domainWhich website influenced the answer
Platform coverageWhere the brand or source appears
Citation recurrenceHow often a source appears across prompts

Why AI Citation Research Matters

Traditional SEO research normally begins with ranking pages and backlinks.

AI citation research adds another layer.

It reveals:

  • Which websites AI platforms trust for a topic
  • Which page formats are retrieved
  • Which directories influence recommendations
  • Which competitors dominate generated answers
  • Whether first-party or third-party sources matter more
  • Where your brand is missing
  • Which sources may be worth targeting
  • Which topics produce retrieval
  • Whether your optimization work changes visibility

If your business is not showing up in AI Overviews, the source list should be a central part of the analysis. The objective is not only to see which brands appear, but also to identify the pages influencing those answers.

The AI Citation Research Workflow

The following process can be completed manually in a spreadsheet or automated through an AI visibility platform.

The underlying research method remains the same.

Step 1: Select the Seed Keyword

Start with one commercially relevant keyword.

Examples:

  • SEO content optimization tool
  • wedding DJ Melbourne
  • cosmetic dentist Austin
  • best accounting software
  • functional medicine London
  • emergency plumber Phoenix

Avoid beginning with an entire list of unrelated keywords.

One seed topic should represent one clear market, product category or service.

Good Seed Keyword

AI SEO tool

This defines a recognizable category.

Weak Seed Keyword

AI

This is too broad. It could produce answers about software, research, careers, models, legislation or general technology.

Record the Context

For each seed keyword, document:

  • Seed keyword
  • Brand name
  • Website
  • Product or service
  • Target audience
  • Country
  • City or service area
  • Main competitors

A practical AI citation research workflow begins with the brand, keyword and website URL. It should also establish whether the business is global, national or local because that context affects the prompts that should be generated.

Step 2: Decide Whether the Search Is Local, National or Global

Location is essential when researching local businesses.

A prompt such as:

Who are the best plumbers?

is not useful without geographic context.

A local prompt should specify the location:

Who are the best plumbers in Austin, Texas?

A national or global product may not require a city.

Local Research Fields

Record:

  • City
  • State, province or region
  • Country
  • Service area
  • Relevant local modifiers

Including the state or region can prevent ambiguity when several cities share the same name.

Example

Brand: Smiles Dentistry
Keyword: cosmetic dentist
Location: Austin, Texas
Website: example.com

This context should remain consistent throughout the research.

Step 3: Expand the Keyword Into Natural-Language Prompts

Natural-language AI prompt workflow.

People do not always use AI platforms like conventional search engines.

Instead of entering:

AI SEO tool

they may ask:

  • What are the best AI SEO tools?
  • Find me an AI SEO tool for a small agency.
  • List the top-rated AI SEO platforms.
  • Which AI SEO tool is the most popular?
  • What AI SEO software should I use?
  • Compare the best AI SEO tools for agencies.

Convert the seed keyword into natural-language prompts that better represent how people interact with AI platforms. Useful variations include “find me,” “list,” “what are the best,” “top rated” and “most popular.”

Five Core Prompt Types

Use at least one prompt from each group.

1. Discovery

Find me the best AI SEO tools.

2. List

List the top AI SEO tools for agencies.

3. Recommendation

What is the best AI SEO tool for a small SEO agency?

4. Comparison

Compare the leading AI SEO tools.

5. Reputation

Which AI SEO tools are the most highly reviewed?

Add Commercial Modifiers

Depending on the topic, add:

  • Best
  • Top rated
  • Most popular
  • Affordable
  • Premium
  • For small businesses
  • For enterprise companies
  • Near me
  • In [location]
  • Alternatives to
  • Compared with
  • Recommended
  • Highest reviewed

Do Not Make Every Prompt Too Specific

A prompt such as:

What is the best cosmetic dentist in Austin for children under 12 who need a specific procedure and offer appointments on Sundays?

may be realistic for one customer, but it creates a very narrow result set.

Use a balance:

  • Broad category prompts
  • Commercial prompts
  • Use-case prompts
  • Location-specific prompts
  • A small number of highly specific prompts

This creates enough coverage to identify patterns.

Step 4: Choose the Platforms to Test

Consider testing the following AI platforms:

  • ChatGPT
  • ChatGPT with web search
  • Google Gemini
  • Google AI Mode
  • Perplexity

These platforms do not behave identically.

A Search Everywhere Optimization strategy must account for the fact that AI platforms’ answers, retrieval sources and market importance can differ.

ChatGPT Without Web Retrieval

ChatGPT may answer from its static training corpus.

When this happens, it may:

  • Provide no citations
  • Mention brands learned during training
  • Use information with a cutoff date
  • Produce a different brand set from web-enabled search

This view helps determine whether a brand may already be associated with the category inside the model’s existing knowledge.

ChatGPT With Web Retrieval

Web-enabled ChatGPT searches for current information and can expose sources.

Use this mode when the objective is to extract citations.

Track standard ChatGPT and web-enabled ChatGPT separately. A standard response may rely on previously learned information, while web-enabled search can provide current and retrievable citation data.

Google Gemini

Gemini can produce brand recommendations and retrieval sources that differ from ChatGPT.

Track it independently rather than assuming the same strategy will work on both.

Google AI Mode

Google AI Mode provides a conversational experience with retrievable sources and a saved chat history.

This makes it different from the standard AI Overview displayed in search.

Perplexity

Perplexity is useful for citation research because sources are central to how its answers are presented.

Still, its smaller usage should be considered when weighting the commercial importance of results.

Step 5: Track Google AI Overviews Separately

Google AI Overviews should not be treated exactly like ChatGPT or Google AI Mode.

AI Overviews differ from chat-based AI products because:

  • The user cannot maintain the same type of conversation
  • The result is connected to a traditional Google SERP
  • The answer may not be saved in a historical chat
  • The result can disappear or change between searches
  • Tracking often requires SERP monitoring rather than chat-history analysis

AI Overviews should therefore be treated as part of traditional Google result monitoring rather than as a standard AI chat platform.

For AI Overviews, record:

  • Keyword
  • Date
  • Device
  • Location
  • Whether an AI Overview appeared
  • Brands mentioned
  • Citation URLs
  • Traditional organic results
  • Local-pack results, when present

Screenshots can also be useful because the answer may not be available later.

Step 6: Force Web Retrieval When Necessary

Not every AI prompt triggers a web search.

A general question may be answered entirely from the static corpus.

To increase the likelihood of retrieval, include a freshness modifier.

Examples:

  • What are the best AI SEO tools right now?
  • What are the best AI SEO tools in 2026?
  • Which AI SEO tools are most popular this month?
  • What are the latest AI SEO platforms as of August 2026?

Adding a freshness modifier such as “right now,” the current year or the current month can increase the likelihood of web retrieval because previously learned information may not be current enough to answer the prompt.

Record Whether Retrieval Occurred

Add a field to your spreadsheet:

Web retrieval used: Yes / No

Do not mix non-retrieval answers with citation-based answers when calculating citation metrics.

An answer cannot provide citation data if no web retrieval occurred.

Step 7: Run Each Prompt Across Every Platform

Create a repeatable testing matrix.

PromptChatGPTChatGPT SearchGeminiAI ModePerplexity
Find me the best AI SEO tools
List top AI SEO tools
Best AI SEO tool for agencies
Compare AI SEO platforms
Most highly reviewed AI SEO tools

Testing several prompts across multiple AI environments creates a broader sample. Use enough coverage to identify recurring brands, sources and patterns within the category.

Keep the Test Conditions Consistent

Use the same:

  • Prompt wording
  • Location
  • Language
  • Date range
  • Brand spelling
  • Target market

Do not change several variables at once.

Otherwise, you will not know why the answer changed.

Step 8: Record Every Brand Mention

For each answer, list the brands in the order they appear.

Example:

PlatformPromptBrandPosition
GeminiBest AI SEO toolsSemrush1
GeminiBest AI SEO toolsAhrefs2
GeminiBest AI SEO toolsRankability3

What Position Means

AI position is not a traditional Google ranking.

It means the order in which the brand appeared inside the generated answer.

AI position should be distinguished from traditional SERP position. Use average visibility and average recommendation position to assess performance across prompts and platforms.

Brand Visibility Formula

A simple manual calculation is:

Brand visibility = Answers mentioning the brand ÷ Total answers × 100

Example:

  • Total answers: 25
  • Answers mentioning the brand: 10

10 ÷ 25 × 100 = 40% visibility

Average Position Formula

Add all positions where the brand appeared, then divide by the number of mentions.

Example positions:

  • 2
  • 3
  • 1
  • 4

2 + 3 + 1 + 4 = 10

10 ÷ 4 = 2.5 average position

Do not assign a position when the brand was not mentioned.

Track visibility and position as separate metrics.

Step 9: Extract Every Citation URL

Open the source or citation section for each retrieval-based answer.

Record every URL, not only the source mentioning your brand.

Your citation sheet should contain:

  • Platform
  • Prompt
  • Date
  • Citation position
  • Full citation URL
  • Root domain
  • Page title
  • Source type
  • Brands supported
  • Whether your brand appears on the page
  • Whether the citation supports the claim
  • Notes

Citation Extraction Table

PlatformPromptCitation URLDomainSource type
PerplexityBest AI SEO toolsexample.com/best-ai-seo-toolsexample.comListicle
ChatGPT SearchAI SEO tools for agenciesreviewsite.com/seo-softwarereviewsite.comReview
AI ModeCompare AI SEO toolsbrand.com/comparisonbrand.comFirst-party comparison

Do not record only the domain.

The individual page matters because two pages on the same website may have completely different levels of relevance and visibility.

Step 10: Verify Every Citation Manually

AI citations are not always perfectly aligned with the response.

An AI platform may associate a citation with a brand even when the brand does not clearly appear on the cited page. This is why every important citation should be reviewed manually.

Open every important citation and confirm:

  1. The page exists.
  2. The content is accessible.
  3. The page discusses the topic.
  4. The cited claim is supported.
  5. The named brand appears where expected.
  6. The content is current.
  7. The page is not merely a homepage.
  8. The citation is not a hallucinated association.

Add a verification field:

Verified: Yes / No / Partial

Why Verification Matters

A high citation count can be misleading when:

  • Several URLs redirect to the same page
  • The AI cites a homepage
  • The source does not contain the brand
  • The cited passage does not support the statement
  • Tracking parameters create duplicate URLs
  • The page is outdated
  • The page has been removed

Raw citations are research inputs, not automatically verified facts.

Step 11: Normalize the Citation Data

AI platforms may return several versions of the same URL.

Examples:

  • https://example.com/article
  • https://www.example.com/article/
  • https://example.com/article?utm_source=chatgpt
  • http://example.com/article

Normalize them into one canonical URL.

Normalization Process

  1. Convert the domain to lowercase.
  2. Remove tracking parameters.
  3. Remove fragments.
  4. Standardize trailing slashes.
  5. Resolve redirects.
  6. Combine HTTP and HTTPS versions.
  7. Separate the root domain from the page URL.
  8. Remove obvious duplicate records.

Without normalization, one source may incorrectly appear to be several different citations.

Step 12: Categorize the Sources

Group each citation by page type.

Useful categories include:

  • First-party website
  • Competitor website
  • Directory
  • Review platform
  • Best-of listicle
  • Comparison page
  • News publication
  • Industry publication
  • Forum or community
  • YouTube
  • Social platform
  • Research paper
  • Government source
  • Homepage
  • Product page
  • Service page

This reveals the type of content influencing the answer.

Example Interpretation

If most citations come from comparison listicles, the strategy may involve:

  • Creating a stronger first-party comparison
  • Getting included in existing listicles
  • Publishing third-party comparison content
  • Earning links to your own comparison page

If directories dominate, prioritize:

  • Complete profiles
  • Correct categories
  • Reviews
  • Internal directory visibility
  • Inclusion on category pages

If first-party competitor pages dominate, study:

  • Their content structure
  • Supporting data
  • rankings
  • Internal links
  • Backlinks
  • Topical coverage

Step 13: Count Citation Frequency

The most valuable source is not necessarily the one appearing in one answer.

Look for recurrence.

URL Citation Count

Count how many times each exact URL appears.

URLCitation count
example.com/best-ai-seo-tools9
reviewsite.com/ai-seo-comparison6
competitor.com/ai-seo-platform4

Domain Citation Count

Combine all URLs from the same domain.

DomainCitation countUnique cited pages
example.com185
reviewsite.com113
competitor.com74

Citation Frequency Formula

For retrieval-based answers:

Citation frequency = Answers citing the source ÷ Total retrieval answers × 100

Example:

  • Retrieval answers: 20
  • Answers citing the domain: 8

8 ÷ 20 × 100 = 40% citation frequency

Prompt Coverage

Count how many unique prompts cite the source.

A domain appearing five times for one prompt is different from a domain appearing across five separate prompts.

Platform Coverage

Count how many platforms cite the source.

SourceChatGPTGeminiAI ModePerplexityCoverage
example.com4
reviewsite.com3
competitor.com2

A source appearing across several platforms may have greater strategic importance than a source appearing repeatedly on only one platform.

Step 14: Separate First-Party and Third-Party Sources

This distinction reveals how the answer is being influenced.

First-Party Source

A page owned by the brand being discussed.

Examples:

  • Homepage
  • Product page
  • Service page
  • Blog article
  • Research report
  • Comparison page

Third-Party Source

A page owned by another publisher.

Examples:

  • Review platform
  • Directory
  • News publication
  • Listicle
  • YouTube video
  • Forum
  • Industry article

Why Third-Party Sources Matter

An AI platform may be more comfortable recommending a business when several independent sources support the same association.

A company’s website says what the company wants people to believe.

Third-party sources provide external confirmation.

Reviews, trusted listicles, YouTube mentions, directories and other independent references can provide important brand-reinforcement signals.

Step 15: Find the Sources Influencing Competitor Mentions

For each leading competitor, identify:

  • How often it is mentioned
  • Its average position
  • Which platforms mention it
  • Which URLs support the mentions
  • Which domains appear repeatedly
  • Whether its own site or third parties dominate
  • Which page types are responsible

Competitor Citation Matrix

CompetitorVisibilityAverage positionTop supporting source
Brand A76%1.8Industry listicles
Brand B60%2.4First-party comparisons
Brand C44%3.1Directories and reviews

This turns AI citation research into an actionable competitive analysis.

Ask the Right Question

Do not ask only:

Why is this competitor appearing?

Ask:

Which sources repeatedly cause this competitor to appear?

That source list becomes the starting point for AI Search Optimization, including content creation, digital PR and targeted outreach.

Step 16: Turn Citation Research Into an Action Plan

Natural-language AI prompt workflow.

Classify every important source into one of four actions.

Action 1: Create

Create content that gets cited by ChatGPT and AI Overviews when the existing citation set reveals a missing first-party asset.

Examples:

  • Original study
  • Cost guide
  • Alternatives page
  • Comparison page
  • FAQ
  • Product benchmark
  • Local data report

Action 2: Improve

Improve an existing page when your site is already cited but performs inconsistently.

Review:

  • Search rankings
  • Direct-answer clarity
  • Original data
  • Internal links
  • Backlinks
  • Freshness
  • Passage structure
  • Page speed

Action 3: Get Included

Pursue inclusion when a recurring third-party source lists competitors but excludes your brand.

Possible actions include:

  • Claiming a profile
  • Updating directory data
  • Requesting editorial consideration
  • Providing expert commentary
  • Supplying original research
  • Building a publisher relationship
  • Offering a legitimate product trial

Action 4: Monitor

Some sources may be authoritative but inaccessible.

A major publisher may not offer a realistic short-term placement opportunity.

Continue monitoring it to understand:

  • What topics it covers
  • Which brands it cites
  • Which evidence it prefers
  • Whether its citation frequency changes

Step 17: Do Not Obsess Over One Prompt

AI answers are unstable.

The same prompt can produce different:

  • Brands
  • Positions
  • Citations
  • Wording
  • Recommendations

Extract and aggregate information instead of treating one generated answer as a fixed result. Each answer represents only one moment in time, while trends across multiple prompts and repeated tests provide more useful information.

Weak Analysis

We ran one ChatGPT search and ranked third.

Better Analysis

Across 25 answers from five prompts and five AI environments, the brand appeared in 36% of responses with an average recommendation position of 3.2.

The second statement describes a pattern.

Step 18: Repeat the Research Over Time

A single extraction creates a benchmark.

Tracking requires repetition.

Use a consistent schedule:

  • Weekly for fast-moving campaigns
  • Monthly for most client reporting
  • Quarterly for broader strategic research

Keep the original prompt set stable.

You may add new prompts, but do not remove or rewrite the original prompts without documenting the change.

Track Movement

Compare:

  • Brand visibility
  • Average position
  • Citation count
  • Unique cited domains
  • First-party citation share
  • Third-party citation share
  • Platform coverage
  • Competitor visibility
  • New source appearances
  • Lost source appearances

Example

MetricJanuaryFebruaryMarch
Brand visibility16%24%40%
Average position4.23.82.9
Unique citations3711
Platform coverage124

This shows whether the campaign is increasing total AI visibility.

A Manual AI Citation Tracking Spreadsheet

Create one tab for prompts and one for extracted answers.

Prompt Tab

FieldExample
Seed keywordAI SEO tool
BrandExample Brand
Websiteexample.com
MarketGlobal
Prompt IDP01
PromptWhat are the best AI SEO tools?
IntentRecommendation
Freshness modifier2026

Answer Tab

FieldExample
Date5 August 2026
PlatformChatGPT Search
Prompt IDP01
Retrieval usedYes
Brand mentionedYes
Brand position3
Citation URLexample.com/best-ai-seo-tools
Citation domainexample.com
Source typeListicle
First or third partyThird party
Citation verifiedYes
NotesBrand listed in comparison table

Domain Summary Tab

DomainCitationsUnique URLsPrompt coveragePlatform coverage
example.com14454
reviewsite.com9243
competitor.com6532

Brand Summary Tab

BrandVisibilityAverage positionPlatformsSupporting citations
Brand A72%1.8528
Brand B48%2.6417
Your brand20%4.125

Common AI Citation Tracking Mistakes

Tracking Only Your Brand

You also need the complete source and competitor landscape.

Otherwise, you cannot identify what is influencing the answer.

Using Only One Platform

A strong result in Gemini does not guarantee visibility in ChatGPT or Google AI Mode.

Treat each AI platform as a separate research environment, even though traditional SEO practices may support visibility across several of them.

Tracking Only One Prompt

One prompt provides too little surface area.

Use several prompt types and at least enough platform combinations to create a meaningful sample.

Forgetting to Force Retrieval

Static answers may contain brand mentions but no citations.

Use current-date modifiers when citation extraction is the objective.

Counting Duplicate URLs

Normalize tracking parameters, redirects and alternate URL formats.

Trusting Every Citation

Open and verify the page.

AI platforms can create weak or inaccurate associations between a source and a claim.

Ignoring First-Party Sources

Third-party mentions matter, but your own pages can still become important retrieval sources when they rank and provide useful evidence.

Ignoring Traditional SEO

SEO is not dead, because retrieval sources normally do not appear from nowhere. The HTML vs JavaScript setup can affect whether a strong search presence leads to a page being discovered and used by AI retrieval systems.

Treating AI Position Like a Google Ranking

AI position represents order within a generated answer, not a stable SERP position.

It must be interpreted across repeated answers.

Frequently Asked Questions

How Many Prompts Should I Track?

Five carefully selected prompts can create a useful starting point when tested across several platforms.

The more commercially important the market, the larger the prompt set should become.

Prioritize quality and coverage rather than producing dozens of near-identical prompts.

How Many Answers Do I Need?

There is no universal minimum number of answers. Start with several distinct prompts across multiple AI environments and continue testing until recurring brands, sources and patterns become clear.

The appropriate sample size depends on the market, number of platforms and importance of the research.

Can I Track AI Citations Without Paid Software?

Yes.

You can run the prompts manually and record the results in a spreadsheet.

The disadvantage is time.

Automation becomes more valuable when tracking:

  • Many keywords
  • Several brands
  • Multiple clients
  • Several platforms
  • Weekly changes

Why Does ChatGPT Sometimes Show No Citations?

The answer may have come from the static corpus rather than live web retrieval.

Use web-enabled search or add a freshness modifier such as the current year.

Should I Count a Brand Homepage as a Citation?

Record it, but classify it separately.

A homepage citation is less useful for content research than a specific article, comparison page or dataset.

When identifying specific source opportunities, filter or classify brand homepages separately from articles, comparisons, studies and other individual pages.

What Is the Most Important AI Citation Metric?

There is no single perfect metric.

For research, focus on:

  • Citation recurrence
  • Prompt coverage
  • Platform coverage
  • Source relevance
  • Verification

For brand performance, add:

  • Brand visibility
  • Average position
  • Competitor share of voice

Can a Citation Appear Without Mentioning My Brand?

Yes.

The platform may use a page to support the general answer without using it as evidence for your business.

Verify each relationship manually.

Does Getting Cited Guarantee Traffic?

No.

The AI Overviews CTR impact is visible when many users read the generated answer without clicking a source.

Citation research is also valuable because it reveals which sources influence visibility and recommendations.

AI Citation Research Checklist

TaskComplete
Seed keyword selected
Brand and website recorded
Market and location defined
Five or more prompts created
Commercial intent represented
Freshness prompts included
Platforms selected
AI Overviews separated from chat platforms
Every prompt tested
Brand mentions recorded
Brand positions recorded
Every citation URL extracted
URLs normalized
Citations manually verified
Sources categorized
First-party and third-party sources separated
Citation frequency calculated
Prompt coverage calculated
Platform coverage calculated
Competitor sources analyzed
Actions assigned to important sources
Next tracking date scheduled

Final Takeaway

Learning how to track AI citations is not simply a matter of asking ChatGPT where it found an answer.

A reliable process requires enough data to reveal patterns.

Start with one seed keyword. Expand it into natural-language prompts.

Run those prompts across the major AI platforms. Force retrieval when necessary.

Extract every source. Verify the pages.

Normalize the URLs. Count recurring domains.

Compare the brands. Repeat the analysis over time.

The brand mentions tell you who is winning today.

The citations help explain why.

That is where the strategy begins.

WRITTEN BY

Zea Gatdula
Zea Gatdula

Zea Gatdula is an SEO consultant who has helped local service businesses grow their organic traffic and rankings for years. She specializes in data-driven strategies that win in Google Search, AI Overviews, and ChatGPT, backed by hands-on client results rather than guesswork.