Starting October 12, 2026, Google began auto-enrolling some Search and Performance Max advertisers — specifically those with linked location assets — into a new feature called “Automated Promotions,” according to Search Engine Roundtable. The system scans your website for anything that looks like a deal and attaches it to your ad as a promotion asset, without a human reviewing the wording first. If you run Google Search Ads or Performance Max for a law firm or professional services business, you are very likely enrolled right now, by default.
What Is Google Actually Turning On Here?
Per Search Engine Roundtable’s October 5, 2026 report, Google’s own advertiser email describes Automated Promotions as a feature “designed to improve your ad’s performance by providing users with valid offers without requiring any manual setup on your part.” In practice, that means Google’s crawler reads your website, looks for language it interprets as a promotion, deal, or coupon, and inserts it into your ad as a promotion asset — automatically, continuously, and without a review step on your end.
Why Should A Law Firm Care More Than A Retailer Does?
A retailer auto-surfacing “20% off” is a pricing question. A law firm auto-surfacing fee language is a regulatory one. Most state bars regulate exactly how firms can advertise fee arrangements, guarantees, and free offers — required disclaimers on contingency fees, restrictions on words like “free” or “guaranteed,” rules on what a “no fee unless we win” claim has to say nearby. None of that nuance travels with an automated scrape. Google’s system is matching patterns that look like offers, not applying your bar’s advertising rules.
What Might Google’s Crawler Flag As A “Promotion” On Your Site?
This is the part most firms miss, because they don’t think of themselves as running promotions at all. Common website language that could read as a “deal” to an automated system includes:
“Free consultation” or “free case review”
“No fee unless we win” or other contingency-fee framing
A flat-fee offer for a specific service (uncontested divorce, simple will, etc.)
Seasonal or limited-time discount language on a services page
Any banner or callout mentioning a reduced rate, waived fee, or percent off
Any of these, pulled verbatim into an ad asset with no disclaimer and no review, is the exposure here — not that Google built the feature, but that it can go live unreviewed.
Is This Opt-In Or Opt-Out — And Did Everyone Get Notified?
Search Engine Roundtable reports that Automated Promotions is opt-out, not opt-in: advertisers are enrolled by default unless they turn it off in account-level Automated Assets settings. The rollout is also phased — only some advertisers have received Google’s notification email so far, which means plenty of accounts are already enrolled with no notice at all. One built-in limit: campaigns that already have promotion assets manually attached are excluded from auto-enrollment, per the same report.
Where Do You Actually Turn This Off?
The control lives at the account level, in Google Ads’ Automated Assets settings — not inside an individual campaign. That’s worth knowing before you go looking for a per-campaign toggle that doesn’t exist. If you or your agency manage Search or Performance Max campaigns with location assets attached, that settings page is where this gets decided, one way or the other.
Should You Panic About This?
No. This is a settings-check story, not a crisis. The feature isn’t malicious, and Google isn’t targeting law firms specifically — it’s a broad ad-performance feature that happens to collide with a regulated industry. The move is simple: go look at your Automated Assets settings, decide deliberately whether to opt in or stay out, and if you do leave it on, check what it’s pulling from your site on a regular basis rather than assuming it’s fine.
Does this apply to Local Services Ads too?
The reporting specifically describes Search and Performance Max campaigns with linked location assets — not Local Services Ads directly. If your firm runs LSAs, it’s still worth confirming with whoever manages your account whether any linked Search or Performance Max campaigns are in scope.
Will an auto-added promotion get my ad rejected, or put my bar license at risk?
There’s no indication this triggers ad rejection on its own. The exposure is advertising-compliance risk if the auto-pulled language runs afoul of your state bar’s rules on fee advertising — not a platform penalty. Check your own state bar’s advertising rules, and talk to your own counsel if you’re unsure where a specific phrase lands. This article is informational commentary, not legal advice.
Can I switch this off for one campaign and leave it on for others?
Based on current reporting, the control is account-level, not per-campaign. If any of your Search or Performance Max campaigns carry linked location assets, the account-wide setting will govern all of them.
What if I’m already running promotion assets on my ads?
Per Search Engine Roundtable, campaigns that already have promotion assets manually attached are excluded from auto-enrollment — so firms that already manage this deliberately are, for now, in the clear on those specific campaigns.
If you’re checking your Google Ads settings anyway, it’s worth checking what’s actually living on your website, too — because that’s the raw material Google’s crawler is reading. SiteScan, a free tool from GrowthAgents, grades five pages of your site, rewrites your weakest page, and shows you a before-and-after score in about two minutes, no card required. It won’t tell you what Google’s ad crawler will do with your copy, but it will tell you what that copy is actually saying — which is the first thing worth knowing either way.
SB 574, signed by Governor Newsom on September 30, 2026, is the first state law written specifically for how attorneys and arbitrators use generative AI — not a bar-association guideline, a statute. It requires attorneys to personally verify every citation in AI-assisted filings, restricts what confidential data can go into AI tools, and bars arbitrators from delegating any part of their decision-making to AI. FindLaw and King & Spalding both flag it as a preview of where AI accountability law is headed nationally.
What Does SB 574 Actually Require?
The law adds a new Business and Professions Code section, §6068.1, and amends California Code of Civil Procedure §128.7. Together they require attorneys to personally verify every case citation and statutory reference in a court filing that used generative AI. A filing with an unverified, inaccurate citation can draw sanctions, and the challenged party gets a 21-day window to withdraw or correct the paper before sanctions proceed. The law firm itself is jointly responsible for violations by its lawyers and staff, absent exceptional circumstances — this isn’t a one-lawyer problem, it’s a firm problem. (FindLaw)
What Can’t Attorneys Put Into AI Tools?
SB 574 bars attorneys from entering confidential, personal, or nonpublic information — dates of birth, medical records, financial details, account numbers — into generative AI unless access is restricted to the attorney and others bound to protect it. That’s a meaningful limit on which AI tools are usable for client work at all, not just a disclosure rule. (King & Spalding)
What Does the Law Require Attorneys to Disclose?
Attorneys must disclose generative AI use to the court for documents submitted to it, and must consider disclosure when AI creates content for the public. King & Spalding notes the scope here is genuinely ambiguous — commentators read it as reaching situations where AI contributed to the substance of the work, not incidental use like AI-assisted proofreading. (King & Spalding) What’s clear: attorneys can’t delegate the practice of law to generative AI. It can assist. It can’t replace judgment. (FindLaw)
Does This Apply to Arbitrators Too?
Yes, and the bar is stricter. Arbitrators cannot delegate any part of their decision-making to AI, and must disclose to the parties when they rely on AI-generated information from outside the record. That’s a direct response to concerns about AI quietly shaping arbitral outcomes without anyone in the room knowing. (FindLaw)
Does This Affect Firms Outside California?
Not legally — yet. But both sources frame SB 574 as a first mover, not an outlier. The open questions commentators are flagging — exactly who must verify a citation, how broad “disclosure” really runs, and whether any of this reaches transactional work rather than just litigation — are the same questions every state will eventually have to answer. (King & Spalding) If your firm already has a clean answer to “who checked this citation” and “what did we let the AI see,” you’re ahead of wherever your state lands.
Does SB 574 apply if my firm isn’t in California?
Not directly. It’s California state law, binding on attorneys and arbitrators practicing there. But it’s the first statute of its kind, and the verification and confidentiality standards it sets are a reasonable preview of what other states will eventually codify.
Does the law cover transactional work, or only litigation?
The citation-verification and sanctions provisions are tied to court filings. The confidentiality restriction and the “can’t delegate the practice of law” language are written more broadly, and commentators note it’s unclear how far they reach into non-litigation work. (King & Spalding)
Is this article legal advice?
No. This is informational commentary on a newly enacted statute, not legal advice. Talk to counsel licensed in your jurisdiction about how SB 574 or any comparable rule applies to your practice.
What should a small firm do right now?
Write down, in plain terms, what your firm already does: who checks a citation before it’s filed, what client data is or isn’t allowed into any AI tool, and who signs off on AI-assisted work before it goes out the door. None of that requires new software — it requires a documented habit. Building it now, before a law like California’s reaches your state, is cheap. Rebuilding trust after a sanctioned filing is not.
Your website makes the same kind of claims your filings do — it just doesn’t get checked as often. Run it through SiteScan, GrowthAgents’ free tool: it grades your site across five pages, rewrites your weakest page, and shows you a before-and-after score in about two minutes. No card required.
By Josh Corbelli, Founder, GrowthAgents AI. Published and last reviewed October 16, 2026.
Answer engine optimization (AEO) is the work of making your firm’s pages easy for ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Claude and Copilot to find, quote and cite. The checklist below has 36 items in 8 groups. Law-firm AI referrals are still small today, near 0.5% of sessions[5], but they are growing fast, and the firms cited now are the firms prospects are shown next.
The uncomfortable fact first: AI assistants send a small share of today’s visits. Three independent datasets put it between 0.17% and 1.08% of site traffic.
Conductor measured 1.08% of sessions from AI referrals across 1,215 enterprise domains and 3.3 billion sessions, May through September 2025[1].
Ahrefs measured 0.17% of visitors on average across 3,000 sites, and 0.56% on sites with fewer than 999 monthly visitors, the size of most small firms (data published February 6, 2025)[2].
Great Jakes, as reported by LawFuel, found AI at 0.47% of sessions on law-firm sites, with 220% growth. The sample size and year were not stated[5].
The case for acting now rests on four numbers, not on current volume.
Growth. Previsible tracked LLM sessions across 166 GA4 properties from 65,249 per month (November 2024) to 644,478 per month (May 2026), a 9.9x increase[3].
Shrinking clicks on the old results. In Pew Research’s March 2025 browsing data, 8% of Google visits with an AI summary produced a click on a traditional result, against 15% without one. Only 1% clicked a link inside the summary[4]. SparkToro and Similarweb put US zero-click Google searches at 68.01% for January through April 2026, up from 60.45% in 2024[6].
Legal queries are exposed. 23.6% of legal queries trigger an AI Overview, and 57.9% of question-style legal queries do (5W Legal AI Visibility Index, April 2026)[7].
Prospects say they will ask AI. In an iLawyerMarketing survey of 1,110 people, 41.9% said they would use ChatGPT to research a lawyer, up from 28.1% in 2025, and 50.1% named any AI engine. This is stated preference, not observed behavior[8].
Visitors who arrive from AI may convert at higher rates, but the evidence is thin. Seer Interactive saw ChatGPT visitors convert at 15.9% against 1.76% for Google organic on one unnamed client site (October 2024 to April 2025). That is one site, and not a law firm[9]. Ahrefs reported AI visitors at 0.5% of its own traffic and 12.1% of its signups[10]. Treat both as directional.
Measurement caveat. Google counts AI Overview and AI Mode clicks inside the “Web” search type in Search Console, so they look like ordinary organic traffic[11]. GA4 can also file ChatGPT visits as “Direct”. Your real AI exposure is larger than the referral report shows.
2. How AI engines choose which firm to cite
An AI engine does not answer your prospect’s question once. It splits it into many sub-queries and cites pages that answer the pieces. iPullRank calls this query fan-out and describes a single prompt expanding into 15 to 20 sub-queries[12]. The example below is illustrative. It was built from that concept and is not measured output from any engine.
Seed query: “best car accident lawyer in Phoenix.” A fan-out could include:
Intent
Possible sub-query
Page that must answer it
Cost
Phoenix car accident lawyer contingency fee percentage
Fees and payment page
Cost
How much does a car accident lawyer cost in Arizona
Fees and payment page
Process
What to do after a car accident in Phoenix
After-an-accident guide
Timeline
How long does a car accident case take in Arizona
Case timeline page
Law
Arizona statute of limitations for car accidents
Deadline page with statute citation
Law
Arizona pure comparative fault car accident
Fault explainer with statute citation
Insurance
Phoenix lawyer for uninsured motorist claims
Uninsured motorist page
Edge case
Rideshare (Uber, Lyft) or truck accident lawyer, Phoenix
One page per claim type
Geography
Car accident lawyer Scottsdale, Mesa or Tempe
Service-area pages with real local content
Language
Spanish-speaking car accident lawyer Phoenix
Language-access section on the firm page
Validation
[Firm name] reviews, Avvo, bar discipline
Reviews, attorney bio, bar numbers
Do not answer a fan-out with 200 near-duplicate pages. Google’s July 10, 2026 guide says mass-producing pages for query variations or fan-out queries violates its scaled content abuse policy[13]. Build one strong page per distinct intent.
3. The 36-item checklist
Each item names what to do and where it belongs. Items marked (Google) come from Google’s published guidance. Items marked (third party) come from iPullRank, Microsoft or the AI vendors’ crawler documentation and are not Google policy.
Group A. Access: can AI systems reach the page? (6 items)
☐ A1. Confirm robots.txt allows Googlebot, and that your CDN or host firewall does too[11](Google).
☐ A2. Allow OAI-SearchBot, the crawler behind ChatGPT search results. Blocking it removes you from ChatGPT search answers[14](third party).
☐ A3. Allow Claude-SearchBot and Claude-User. Blocking them can reduce visibility in Claude’s user-directed search[15](third party).
☐ A4. Allow PerplexityBot and allowlist its published IP list in your firewall[16](third party).
☐ A5. Decide GPTBot and ClaudeBot (model training) separately from search crawlers. OpenAI documents that the two settings are independent[14].
☐ A6. Confirm every key page is indexed, snippet-eligible, and carries no stray nosnippet or noindex tag[11](Google).
Group B. Structure: can a passage stand alone? (6 items)
☐ B1. Open every page and every H2 section with a direct answer of one to three sentences, then support it[17](third party).
☐ B2. Write each paragraph around one concept so it still makes sense when lifted out of the page[12](third party).
☐ B3. Use headings shaped like the query a client would type, such as “How much does a car accident lawyer cost in Arizona”.
☐ B4. Make each heading describe the text under it. Do not use cute or vague headings.
☐ B5. Put fees, timelines and eligibility in tables or lists, in HTML text, never inside images, PDFs, tabs or accordions[18](third party).
☐ B6. Keep headings and paragraphs clean and semantic. Google says perfect code is not required[13](Google).
Group C. Content: is it worth citing? (6 items)
☐ C1. Add first-hand detail only your firm has: case types handled, local court practices, intake process. Google calls this “non-commodity” content[13](Google).
☐ C2. Replace vague quantifiers (“most clients”, “many cases”) with numbers and dates.
☐ C3. Cite the statute, rule or court by name and link to the primary source, on every legal claim.
☐ C4. Publish original data from your own practice, such as timelines you can substantiate, in aggregate and without client identifiers.
☐ C5. Quote an attorney in the body of each guide, with name, bar admission and a link to the bio[19](third party).
☐ C6. Add a visible full date (“Last reviewed October 16, 2026”) in the body, not only in page metadata[12](third party).
Group D. Entities and trust (4 items)
☐ D1. Name entities precisely: “Maricopa County personal-injury attorney”, not “attorney”[17](third party).
☐ D2. Give each attorney a bio page with bar number, admissions, education and a headshot. Show the author’s background on each guide[20](Google).
☐ D3. Keep your firm name, address and phone identical on your site, Google Business Profile and directories[21](third party).
☐ D4. Check every claim, superlative and testimonial against your state bar’s advertising rules. ABA Model Rule 7.1 bars false or misleading communications, and states vary[22].
Group E. Structured data (5 items)
☐ E1. Add LegalService JSON-LD with areaServed, address, telephone and openingHours. schema.org marks Attorney as deprecated in favor of LegalService[23].
☐ E2. Add Person markup for each attorney, linked to the firm with worksFor and to profiles with sameAs.
☐ E3. Add Article markup with author, datePublished and dateModified to every guide.
☐ E4. Make every marked-up fact match visible page text, and validate in Google’s Rich Results Test[11](Google).
☐ E5. Treat FAQ markup as optional. Google stopped showing FAQ rich results on May 7, 2026. The markup stays valid, but Google Search value is none[24].
Group F. Local and reviews (3 items)
☐ F1. Keep your Google Business Profile complete and current: categories, services, hours, photos[13](Google).
☐ F2. Build one service-area page per city with real local content, not swapped city names[21].
☐ F3. Respond to every review, within bar-rule limits on confidentiality.
Group G. Linking (2 items)
☐ G1. Link each guide to its related practice-area, attorney and fee pages with descriptive anchor text, never “click here”[17](third party).
☐ G2. Link out to primary sources (statutes, courts, bar rules) so each claim can be checked.
Group H. Measurement and maintenance (4 items)
☐ H1. Verify the site in Search Console and review the Generative AI performance report[13](Google).
☐ H2. In GA4, build a channel group for chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com referrers.
☐ H3. Test 20 fixed prompts monthly in each engine and log whether your firm is cited, quoted or absent.
☐ H4. Re-review every guide every 90 days and update the visible date only when the content changes.
4. What Google says you do not need
Several popular AEO tactics are not required by Google. Google’s July 10, 2026 guide says you do not need new machine-readable files, such as an llms.txt file, and that there is no requirement to break content into tiny pieces or to write in a special way for generative AI search. It also says structured data is not required for generative AI features[13].
We still apply the structure and schema items above. They are third-party practice (iPullRank, Microsoft) and low-cost, and ChatGPT, Claude and Perplexity are not Google. They are hypotheses with supporting guidance, not guarantees. No one, including us, can promise a citation.
5. Why the checklist is harder than it looks
Each item is simple. Doing all 36, correctly, on every page, and keeping them correct, is the hard part.
Crawler access sits in four places: robots.txt, your CDN, your host’s firewall and your security plugin. A single misconfigured layer blocks an AI crawler while every audit tool reports a pass.
Schema must match the page text exactly. A changed phone number or a retired attorney breaks the match across LegalService, Person and Article markup, your Google Business Profile and every directory.
Fan-out coverage is a research job. You need the 15 to 20 sub-queries behind each practice area, then one page per distinct intent, written without creating the duplicate pages Google penalizes.
Legal accuracy is non-negotiable. Statutes change and bar advertising rules differ by state. Every page needs an attorney review.
Measurement is unreliable by default. AI visits hide in Direct traffic and in Google’s Web totals, so you need your own prompt-testing log.
The engines disagree. ChatGPT’s share of AI referrals ranges from 75.96% to 95.1% depending on the panel and month[3], so a tactic validated in one engine is not validated in the rest.
A small firm with 1 to 25 attorneys has no one whose job this is. That gap is where we work.
6. How GrowthAgents and Agency 310 implement it
GrowthAgents AI scores your pages against our AEO rubric, and Agency 310, its exclusive agency implementation partner, does the implementation for you.
Score first. The free GrowthAgents SiteScan grades your site across 5 pages and shows a before-and-after AEO score on one rewritten page.
Implement with Agency 310.Agency 310 builds and delivers cost-effective AI solutions for law firms to manage and grow their practices. The team brings 40+ years of combined small-law-firm business growth experience, AI implementation, custom AI app development, and CRM automation and integration for Clio, HubSpot, Lawmatics and other systems.
Scope. Fan-out query research and revision, schema review and updates, Google Business Profile optimization with local content, crawler-access fixes, and page rewrites against the rubric.
“Most firms fail the checklist at the boring layers: a blocked crawler, a stale phone number in the markup, a guide with no date. We find those first.”Josh Corbelli, Founder, GrowthAgents AI.
AEO is the practice of structuring your site so AI assistants and AI search features can find, quote and cite your firm’s pages when prospects ask legal questions.
Is AEO different from SEO?
The base is the same. Google says pages must be indexed and snippet-eligible to appear in AI features, and that there are no additional technical requirements for Google’s AI Overviews[11]. AEO adds passage-level structure, query fan-out coverage and access for non-Google crawlers.
Do I need an llms.txt file?
Google says no. Other engines have not published support for it, so treat it as optional and low-cost.
How do I know if ChatGPT is sending visitors?
Create a GA4 channel group for AI referrer domains and run a monthly prompt test. Some visits arrive without a referrer and show up as Direct, so the GA4 count is a floor.
Should I use FAQ schema?
It is optional. Google stopped showing FAQ rich results on May 7, 2026. Visible Q&A sections still help readers and may help other engines.
How long does it take?
Crawler and schema fixes can be completed within days. Citation changes depend on how often each engine re-crawls, and no vendor can promise a date.
