How AI Search Works For Legal & Automotive

A Mental Model You Can Use To Understand and Diagnose How Ready Your Business Is For AI Search

You can use this mental model to be able to better understand how you can prepare your business for the new age of AI search. This is the exact same model that I use as our lead AI Search Strategist, and it’s the same model we teach to our team members internally. 

The exact model will differ by platform, but the practical model is similar:

1. The system needs to access information.

2. It needs to understand and organize that information.

3. It interprets the user’s question.

4. It searches beyond the literal words typed.

5. It selects sources or evidence.

6. It generates an answer.

7. It may cite, mention, recommend, or ignore certain businesses.

8. The result may vary depending on the user, location, account state, and timing.

Here is the framework that you can use to understand HOW AI search works, and how to use it to diagnose where the gaps in your current AI search strategy are. 

Note: If you’re lazy like me and you want us to do this for you, contact us and ask for me (Jack Lightner) and I’ll diagnose where the gaps are in your AI search strategy for free if you’re a qualifying law firm or franchised automotive dealership. 

Now without further adieu, here’s the process I use. 

Layer 1: AI Systems Can Only Show Your Customers What They Can Reach

AI search can only mention your dealership or law firm if it has access to information sources about your business. 

This information typically comes from one of three sources. 

1. Its training data;

2. Your website; 

3. Other websites / properties in the search indexes

You have little control over the training data, but you do have control over your website and some control over what the other properties / websites in the search index say about you. These are resources like:

• Citations / Directories (GBP, Bing, Better Business Bureau, FindLaw, etc)

• News articles 

• Forum discussions (Reddit, Public Facebook Groups, etc) 

• Other websites 

If important information is blocked, missing, outdated, buried in images, locked behind forms, or inconsistent across the web, the system has less to work with. You also want to make sure your website content is not dynamically displayed with javascript, as most AI tools cannot currently parse javascript.

For car dealerships, this means your inventory, service offerings, specials, location details, reviews, business hours, OEM affiliations, and department pages need to be visible and current.

For law firms, this means practice areas, attorney bios, service areas, credentials, case-type pages, FAQs, reviews, and trust signals need to be visible and current.

How To Check: 

The easiest way to check what information AI search can surface about your brand is to use individual prompts. Here are some of the prompts I use internally when I am auditing client sites. 

Automotive: 

• Access the website [YourURL.com]. What brands does this store sell? 

• Browse [YourURL.com] and find their active inventory. Extract the exact VIN and price of the most expensive vehicle currently listed on their site. 

• Look at the special offers or lease deals page on [YourURL.com]. What is the expiration date listed for the current monthly specials? 

Legal:

• "Access the website [YourURL.com]. What are the exact names of the partners listed on their team page, and what are their specific practice areas? 

• Based strictly on the disclosures or disclaimer page on [YourURL.com], what is their stated policy regarding attorney-client privilege for online contact forms?" 

• Check the blog or news section of [YourURL.com]. What is the title and publication date of the most recent article published there? 

If these prompts are surfacing the correct information about your business, you can be reasonably sure that your information can be found, and you can move on to the next step. 

Practical takeaway: if the web cannot clearly see it, AI search will not use it. 

Layer 2: Your Information Needs To Be Usable, Not Just Reachable

Just because your information is reachable does not mean it’s usable for how AI search works. AI search does not look at your website the way a human does. Large language models rely on probabilistic weights. When your businesses’ site uses vague language like “we care about our clients” or “we offer great service” or “we give the best service” AI tools assign it a low confidence score. 

Good pages that AI tools like to cite & recommend are pages where:

• They clearly define the page's purpose, what it’s for, why it’s important, and what to do next. 

• They’re dense with named entities (people, places, specific product specifications, pricing, etc). 

• Any claims are backed up with evidence. For example, if you claim to be the best personal injury lawyer in Baton Rouge, you’d want to include your win rate stat, award entity that recognized you as this, number of $ won for your clients in said area, etc. 

• They cite specific data. For example, your car dealership’s warranty page might cite the specific amount of claim dollars paid out for warranty repairs, or another specific state about how much it saves the customers. 

For a dealership, useful information might include current inventory details, brand specialization, service capabilities, warranty information, financing options, customer reviews, trade-in process, and local market context.

For a law firm, useful information might include case types handled, jurisdictions served, attorney credentials, process explanations, common client questions, fee structure basics, and evidence of experience.

How To Check: 

The easiest way to check the quality of your information is not to use some tool, it’s to roll up your sleeves, grab a highlighter, and sit down. Pull up the page you want to review and then run it through this test that I use for product & services pages in automotive and legal. Pull up a section of ~300 words. Imagine you can’t see the rest of the website. 

Automotive Example: New Vehicle Specials Page

1. [WHAT] (Yellow): Brand & Trim Architecture. Highlight exact years, makes, models, configurations, or exact parts/packages (e.g., “2026 Ford F-150 Lightning XLT Extended Range”). Do not highlight general phrases like "our great selection of trucks."

2. [WHERE] (Green): Franchise Radius & Logistics. Highlight specific physical addresses, local cities, or specific delivery bounds (e.g., “serving Haverhill and the Merrimack Valley within a 25-mile radius”).

3. [WHO] (Blue): Financing & Incentives Tier Qualification. Highlight exact eligibility limits (e.g., “Tier 1 credit approval required via Ford Credit,” “active-duty military personnel,” or “current lessees of non-GM vehicles”).

4. [HOW MUCH] (Pink): Transactional Math. Highlight exact monetary figures or contract limits (e.g., “MSRP $54,200,” “$399/month for 36 months,” “$4,500 due at signing,” or “10,000 allowable miles per year”).

5. [WHY] (Orange): Verifiable Validation Entities. Highlight factual proof anchors (e.g., “CARFAX 1-Owner history report,” “Ford Gold Certified Pre-Owned 12-Month/12,000-Mile Warranty,” or “4.8-star Google rating across 1,200 reviews”).

6. [HOW] (Purple): Low-Friction Conversion Action. Highlight the explicit programmatic next step (e.g., “Use our online Trade-In Appraisal tool,” “Click Pre-Qualify for Financing,” or “Schedule a service bay online”).

Legal: 

1. [WHAT] (Yellow): Specific Matter Niche & Scope. Highlight the granular legal sub-type and what is legally covered or excluded (e.g., “Commercial Trucking Jackknife Collisions,” “Chapter 11 Business Reorganization,” or “Uncontested Marital Dissolutions”). Do not highlight vague copy like "aggressive legal advocates."

2. [WHERE] (Green): Active Jurisdictional Authority. Highlight the exact courts, bars, or districts where the firm actively files cases (e.g., “licensed in Massachusetts Supreme Judicial Court,” “U.S. District Court for Eastern Michigan,” or “representing clients in Essex County Circuit Courts”).

3. [WHO] (Blue): Targeted Client Thresholds. Highlight clear qualification prerequisites (e.g., “individuals denied Workers' Compensation benefits after a workplace injury,” or “privately held companies with annual revenues over $2M facing patent letters”).

4. [HOW MUCH] (Pink): Financial Fee Structure. Highlight the baseline structural terms of the engagement (e.g., “Contingency fee of 33.3% prior to filing a lawsuit,” “Flat-rate contract review of $1,200,” or “Hourly billing at $450/hour with a $5,000 initial retainer requirement”).

5. [WHY] (Orange): Publicly Substantiated Records. Highlight verifiable proof markers, specific case data, or peer metrics (e.g., “Docket No. 24-CV-1082,” “$1.2M settlement in Middlesex County,” or “Super Lawyers inclusion from 2021 to 2026”).

6. [HOW] (Purple): Compliance Engagement Protocol. Highlight the direct next step to establish contact (e.g., “Complete our secure 5-minute online conflict-check form,” or “Call our intake team for a confidential 15-minute case case triage”).

Once highlighted, calculate your AI Usability Score for each 350-word chunk using this simple manual math step:

[ Number of HIGHLIGHTED words in the chunk ] 

--------------------------------------------  x 100  = Usability Score 

[ 300 Total Words in the chunk ]

The AI Usability Scorecard

• 70% to 100% (High Density / AI-Optimized): Excellent. The chunk is tightly packed with structured facts. AI search engines can easily map this to user intents, extract information, and confidently use it as a cited snippet.

• 40% to 69% (Functional / Borderline): Passable. The data is present, but it is surrounded by unnecessary text. It risks being filtered out if a competitor's page states the facts more cleanly.

• 0% to 39% (Unusable Fluff): Critical Failure. If a chunk scores below 40%, it means most of the words are filler adjectives (“trusted,” “premier,” “dedicated,” “affordable”). An AI engine's retrieval process will score this chunk poorly for similarity and entirely skip it when answering user questions.

Long story short, you want to aim for your content to score at somewhere between a

Practical takeaway: AI systems need clean, structured, specific information they can confidently use that’s NOT layered with fluff. 

Layer 3: What The AI Searches For Is More Than What Was Typed

The prompt you put into the AI tool is NOT ultimately what the AI searches for. AI Search tools utilize a process called the query-fan out. 

A query fan-out is an AI search technique that takes a single, simple user prompt and breaks it down into multiple, highly specific background searches. Instead of just searching the exact words you typed, the AI acts like a human researcher. It thinks about all the different pieces of information needed to give you a perfect answer, and it searches for all of those pieces at the exact same time.

For example, take the search “best toyota dealership near me”. Let’s assume you live in Houston, TX. A query fan out in this case would want to break down the question into smaller questions to help you find a coherent answer. This is the same process that a human researcher might make multiple searches to help figure out which Toyota dealership is “best”. The AI might search things like:

1. Geolocation & Region-Specific Queries

• Toyota dealerships in Houston TX map

• Toyota dealers North Houston I-45

• Toyota dealerships Katy Freeway West Houston

• Toyota dealers near Sugar Land Pearland TX

2. Reputation & Volume Rankings

• Best rated Toyota dealership Houston TX Google reviews

• Toyota Presidents Award winners Houston Texas

• Top 10 Toyota dealers in Houston customer service rating

• Reddit Houston best Toyota dealer to buy from

3. Inventory & Pricing Strategy Queries

• Largest Toyota inventory Houston TX

• Certified Pre Owned Toyota dealer Houston

• Toyota truck deals Tundra Tacoma Houston TX

• Houston Toyota dealerships no dealer markup reviews

4. Local Competitor Identification

• Toyota dealership #1 Houston reviews

• Toyota dealership #1 vs toyota dealership #2

• Toyota dealership #2  Houston TX customer complaints

• Toyota dealership #2 service center rating

The new era of AI search is about covering this query fan-out in as much depth as possible. The better you’re able to cover the set of queries that go along with your most common searches, the more likely you are to be recommended, cited, and win business from AI search. 

How to check: 

To check which query fan-outs are important to your business, you must map your high-value conversion pages hidden paths to ensure your brand is retrieved when AI platforms decompose a user's intent. That sounds complicated, but it’s easier than you think. 

Step 1: Extract Hidden Fan out Queries 

Because AI search engines do not publish an official keyword list for their internal sub-queries, you have to extract them yourself. We use an internal tool to do this, but you can effectively do the same thing with tools like:

• OtterlyAI’s Query Fan Out Analysis Tool (Free w/email)

• Wellow Query Fan-out Generator (Free)

These tools will give you a list of related searches for you to work through. 

Step 2: Prioritize By Importance

Not every sub-query generated by an LLM is worth your time. Filter your discovered fan-outs by applying the same four-step prioritization framework we use internally. To fill this out, create a spreadsheet with your discovered fan-out queries in rows, and these four questions as columns. Score or categorize each query using the criteria below. 

1: Does the fan out query have commercial intent?

AI engines break broad prompts down into highly specific sub-queries. Filter out purely informational or educational queries unless they directly lead to a sale. Informational queries should be the last set you hit as they tend to have the lowest total impact on your businesses’ bottom line. 

How to check: Look for transactional and investigational modifiers within the fan-out string.

• Low Commercial Intent: "What is...", "History of...", "How does [X] work". (Discard or deprioritize).

• High Commercial Intent: "Best [dealership] for [specific use case]", "[lawyer #1] vs [lawyer #2]", "Does [dealership] have zero down lease offers for  [model]". 

We recommend that you start with all high commercial intent queries first. Make sure you have topical coverage for these commercial queries, and then work up to more informational searches. 

2: Is the fan out query something you can address onsite, or offsite?

AI models do not just search your website; they search the whole web to answer a prompt. You must categorize where the fix needs to live.

• Onsite Queries: These ask for authoritative facts about your specific product, features, pricing, or documentation.

  • Example: "Does [Your Brand] have an open API?" 

    • Action: Onsite fix - update FAQs or create a page as needed. 

• Offsite Queries: These ask for unbiased, third-party validation, reviews, or social proof.

  • Example: "Is [Your Brand] reliable according to Reddit?" or "What are the complaints about [Your Brand] on G2?" 

    • Action: Offsite fix. Work on PR, do outreach to candidate sites represented in the organic results for the fan out. 

3: Have you covered the fan-out query in detail already?

This is your content gap analysis. Check if your current digital footprint actually answers the sub-query clearly enough for an AI to extract it.

Fully Covered: The answer exists in a clear, structured format (like an H2 heading + a bulleted list) on your site or active third-party pages.

Partially Covered: The information is buried deep inside a 3,000-word blog post or PDF manual where an LLM might miss it or give it low weight.

Not Covered: You have completely ignored this specific angle, feature, or comparison.

4: How effortful will it be to cover this fan-out? 

Rate the resource cost to execute the fix so you can target "low-hanging fruit" first.

• Low Effort: Requires a simple copy tweak, adding an FAQ schema block to an existing page, or replying to an active forum thread.

• Medium Effort: Requires writing a new dedicated landing page, creating a comparison table, or updating product documentation.

• High Effort: Requires launching a full digital PR campaign, restructuring your entire site navigation, or changing your pricing transparency.

Generally, low-hanging fruit for this strategy will involve updating your website first, and then moving to third-party sources outside of your website next. 

Step 5 (Bonus!): Take On & Off Site Action 

Your goal here is to make your content Extractable and Fetchable for AI bots.

• Use Literal Headers: If the fan-out query is "Is [Product] safe for enterprise use?", create an <h2> tag on your security page that says exactly: <h2>Is [Product] Safe for Enterprise Use?</h2>.

• Feed the LLM in Bullets: Directly below that header, provide a 2-3 sentence direct summary, followed by a bulleted list of facts. LLMs prioritize tables and bullet points for synthesis.

• Deploy Schema Markup: Use Product, FAQ, and Organization schema. This structures your data in a standardized format that AI crawlers read instantly without needing to guess the context. 

Offsite Action Plan

Your goal here is to influence the third-party sources the AI trusts for unbiased answers.

• Forum Seeding (Reddit/Quora): If AI frequently fans out to Reddit to check user sentiment, ensure your team or community managers are actively answering questions on relevant subreddits. Optimize those responses with clear brand mentions. Note: Don’t do this blindly. You will get banned and your posts will get removed. 

• Review Platform Optimization: For auto/legal, LLM fan-outs heavily scrape sites like BBB, DealerRater, FindLaw, Avvo, and others. Direct your customer success teams to ask users for reviews that mention the specific features discovered in your fan-out analysis. It’s also very helpful to seed certain phrases in your reviews. 

• Digital PR & Citations: If the fan-out looks for "Top lawyers for [X]", pitch industry publications and blogs that already rank for that sub-query to get your brand added to their lists.

Practical takeaway: optimize around the individual information points that a customer uses to make a decision, not just an individual search phrase.

Layer 4: Selected Sources Are Not Identical To Organic Search Rankings

AI search still depends heavily on traditional search systems. Google has said its generative AI search features are rooted in core Search ranking and quality systems.

But the sources selected for an AI answer are not always the same as the top organic links.

Academic research on Google AI Overviews has found that AI-cited domains can differ from the co-displayed first-page search results. That means businesses should not assume that ranking well automatically means being cited, or that not ranking #1 means they cannot be included.

For dealerships and law firms, this creates both risk and opportunity.

The risk: a competitor, directory, marketplace, review site, or third-party article may shape the answer more than your own website.

The opportunity: if your information is clear, useful, fresh, and reinforced across trusted sources, you may be included in AI answers even when the classic ranking picture is more competitive.

Practical takeaway: rankings still matter, but AI visibility is broader than rankings.

Layer 5: Citations Don’t Guarantee Truth Or Accuracy

One of the biggest mistakes decision makers can make is assuming that an AI citation means the answer is correct.

AI systems can cite a source and still summarize it incorrectly. They can omit context. They can overgeneralize. They can pull outdated information. They can use sources that are not the best available sources.

That matters for lawyers and dealerships because accuracy affects trust.

If an AI says your dealership does not offer a service you actually offer, that is a business problem.

If an AI misstates a law firm’s practice area, location, fee model, or attorney credentials, that is a business problem.

If an AI recommends competitors because their information is easier to find or better reinforced across the web, that is a business problem.

Practical takeaway: AI visibility is not just about being present. It is about being represented accurately.

Layer 6: Results Are Personalized And Contextual

Two people can ask the same question and see different answers.

Why?

Because AI and search systems can consider location, language, device, search history, account state, past conversations, memory settings, and the context of the current session.

A logged-out user in one city may see a different answer than a logged-in user in another city.

A shopper who has been asking about trucks may get different recommendations than someone who has been asking about compact SUVs.

A legal consumer searching from one county may see different options than someone searching from another.

Practical takeaway: one screenshot is not the truth. It is one observation.

Layer 7: Variation Is Natural

AI results can change from one run to the next.

That does not always mean something is broken. These systems are dynamic. They may use fresh sources, different query paths, different ranking systems, model updates, experimentation, or different levels of randomness.

That is why AI search measurement needs to be repeated over time.

A serious visibility program should track patterns, not one-off answers.

For example:

• Are we mentioned across many relevant category entry points?

• Are we cited more or less often over time?

• Are competitors being recommended more often?

• Are the answers accurate?

• Are the sources supporting the answers credible?

• Are we visible across Google, ChatGPT, Perplexity, Copilot, and other surfaces?

Practical takeaway: measure share of answer, not just a single answer.

Layer 8: Optimize For Visibility, Not Just Rankings

The goal is not to “trick” AI systems.

The goal is to make your business easier to understand, verify, compare, and recommend.

That means the work includes SEO, but it also includes content strategy, reputation management, local data accuracy, review generation, third-party visibility, product/service data, digital PR, and measurement.

For dealerships, that may mean better inventory data, stronger model pages, clearer service pages, review velocity, local proof, offer freshness, and marketplace consistency.

For law firms, that may mean stronger practice-area pages, better attorney bios, clearer case-type education, local authority, review strategy, legal directory accuracy, and thought leadership on the questions clients actually ask.

Practical takeaway: AI search rewards businesses that are clearly understood and broadly validated.