No. Google's AI Overview, like ChatGPT and Perplexity, quotes or paraphrases the passage that answers the query and links the page. A case-result disclaimer in a footer, sidebar or separate block does not travel with it. The only qualifier that reliably reaches the reader is one written into the same sentence as the result.
Firm AEO's bar rules post stated this principle and the Canadian rules guide applied it to Model Code Rule 4.2-1. This post adds the US mechanics both left out: what the ABA, New York, Florida and Texas rules ask of a results statement, before-and-after sentences, and how to check what engines say.
- Nothing obliges Google's AI Overview, ChatGPT or Perplexity to repeat a disclaimer that sits in a footer, a sidebar or a separate paragraph. An engine lifts the passage that answers the question and cites the page.
- New York's 22 NYCRR 1200.7.1(e)(3) requires the words 'Prior results do not guarantee a similar outcome' with any advertised past result. The required wording is fixed, so it can be written into the results sentence itself.
- The EBU and BBC study of more than 3,000 assistant answers, published October 2025, found 31% had serious sourcing problems and 20% had major accuracy issues, so a firm cannot assume the engine reproduces its wording faithfully.
- Firm AEO's working rule: every sentence that states a dollar figure, a verdict or a win rate must still be compliant if an engine quotes it with nothing around it.
Why doesn't the disclaimer travel with the quote?
An answer engine retrieves passages, not pages. Google's AI Overview, ChatGPT search and Perplexity each select the few sentences that answer the query, restate them and attach a link. A qualifier placed at the bottom of a results page is a separate passage that answers nothing, so it is rarely the passage selected. The reader sees the number without the condition.
Few readers follow the link to find the condition. Pew Research Center's July 2025 analysis of 68,879 searches by 900 US adults found users clicked a link inside the AI summary in 1% of visits where one appeared. Clicks on any traditional result fell to 8% when a summary appeared, against 15% without one. Most readers never reach the page where the footer sits.
Dean Cook's January 5, 2026 Harvard JOLT Digest piece puts the risk in professional-responsibility terms: summaries 'may omit important conditions, exceptions, or jurisdictional limits' while sounding confident. A lawyer controls an AI Overview only at the sentence level.
What do the bar rules require of a case result?
ABA Model Rule 7.1 does not ban case results. Comment [3] says a truthful report of results may still mislead if it leads a reasonable person to expect the same result 'without reference to the specific factual and legal circumstances' of each case, and that an appropriate disclaimer or qualifying language may preclude that finding. Some states go further and fix the wording or the conditions.
The table is a marketing summary, not legal advice. The firm's ethics counsel decides how each rule applies to a given page.
| Regime | What it says about past results | What that means for the extractable sentence |
|---|---|---|
| ABA Model Rule 7.1, comment [3] | Truthful results can mislead if presented without reference to each client's circumstances; qualifying language may preclude a finding | The qualifier works only if it is attached to the result the engine lifts |
| New York, 22 NYCRR 1200.7.1(d) and (e)(3) | Statements likely to create an expectation about results require the disclaimer 'Prior results do not guarantee a similar outcome' | Fixed wording, short enough to sit in the same sentence or the sentence directly after |
| Florida, Rule 4-7.13 | References to past results are inherently misleading unless objectively verifiable; comment flags atypical results and omitted facts such as a default judgment | The sentence needs the verifiable facts (matter type, year, whether contested), not only a disclaimer |
| Texas, Rule 7.01 commentary | Same unjustified-expectation test and the same note that a disclaimer or qualifying language may preclude a finding | Texas also requires filing with the Advertising Review Committee, so the revised sentence is the version on file |
How do you write the qualifier into the extractable sentence?
Put the result, its circumstances and the qualifier in one sentence, in that order, so the smallest unit an engine can lift still carries all three. Name the matter type, the year and the court or forum, state whether the result was a verdict, a settlement or a default, and end with the qualifier the firm's state requires. The examples below are illustrative, not drawn from a real firm.
The pattern also helps citation. A dated, specific sentence is the kind of self-contained passage engines prefer to quote, while 'millions recovered' gives an engine nothing to cite and cannot be verified.
| Before (fails if quoted alone) | After (compliant if quoted alone) | What changed |
|---|---|---|
| We won $2.4 million for a crash victim. | In 2024 a Kings County jury returned a $2.4 million verdict for a client injured in a rear-end collision; prior results do not guarantee a similar outcome. | Year, forum, verdict vs settlement, New York disclaimer in the same sentence |
| Over $50 million recovered for our clients. | Between 2018 and 2025 the firm resolved 140 injury matters through settlement or verdict; each result depended on the facts of that case. | Unverifiable aggregate replaced by a countable record with a qualifier |
| Charges dropped in DUI case. | In 2025 prosecutors in Hillsborough County dismissed a DUI charge against a client after a suppression motion; the outcome turned on the facts of that stop. | Florida's objectively verifiable standard: who acted, where, why |
| We win 98% of our cases. | Deleted. A win rate needs a definition, a denominator and a period, and few firms can substantiate one under Rule 7.1. | Removed rather than rewritten |
Will the engine still change the wording?
Sometimes, and the published error rates are high enough to plan for. The EBU and BBC News Integrity in AI Assistants study, published October 2025, had journalists from 22 public service media organizations in 18 countries review more than 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity. 45% had at least one significant issue, 31% serious sourcing problems and 20% major accuracy issues.
The BBC's earlier February 2025 round is closer to the quote problem. Of 62 answers that quoted BBC articles, 8 quotes (13%) were altered or absent from the cited article. Both studies tested news content rather than law firm pages, and no published study yet measures how often an AI Overview drops a legal qualifier. Read them as evidence that paraphrase drift is common, not as a legal-specific rate.
A qualifier inside the sentence raises the chance it survives a paraphrase; it does not guarantee it. That is why the next step is checking what the engines actually say.
How do you check what AI engines say about your results?
Run the questions a prospect would ask about the firm's track record, logged out, in Google's AI Mode and AI Overviews, ChatGPT and Perplexity, and record every sentence that mentions a dollar figure, verdict or success rate together with the cited URL. Any figure that appears without its circumstances points to the page that needs rewriting first.
- Prompts to run: '[firm name] verdicts', '[firm name] settlements', 'has [firm name] won cases like mine', and the practice-plus-city prompt for the firm's main practice.
- Record the engine, the date, the exact sentence and the cited URL, using the same ledger fields as the firm's monthly mention tracking.
- Check directory profiles and old press releases, which often carry unqualified figures engines still cite.
- Keep a copy of every revised page with its date. New York requires retention of computer-accessed advertising, and Texas requires filing, so the version history is a compliance record as well as a work log.
Should a law firm keep publishing case results at all?
Most firms can, if every results sentence passes the quoted-alone test and the firm's ethics counsel signs off on the wording. Firm AEO's view is that the risk sits in unqualified aggregates and win rates, not in individual, well-described matters. A firm that cannot describe a result fully, because of client confidentiality under Rule 1.6 or a missing record, should leave that result off the page rather than publish a shorter version an engine can strip further.
Is a results disclaimer in the website footer still worth keeping?
Yes. It still serves readers who reach the page and some state rules may expect it, but it is a backstop. The qualifier that matters for AI answers is the one inside each results sentence.
Is a law firm responsible if an AI Overview drops the qualifier from a compliant sentence?
No bar authority Firm AEO has found answers this directly as of October 2026. The Harvard JOLT Digest commentary of January 2026 notes that the advertising rules were not written for algorithmic intermediaries. What a firm can show is that its own published sentence was compliant on its own, and that it monitors and corrects what it controls. The firm's ethics counsel decides how to handle a specific inaccurate summary.
Does the New York disclaimer have to use the exact wording?
22 NYCRR 1200.7.1(e)(3) sets the wording as 'Prior results do not guarantee a similar outcome.' Because the text is fixed and short, Firm AEO writes it into or directly after the results sentence rather than paraphrasing it. Whether a particular page is an advertisement under the New York rules is a question for the firm's ethics counsel.
Can an engine quote a client testimonial that mentions a result?
Yes, and a testimonial that names a dollar figure raises the same unjustified-expectation question as a results sentence. New York lists client testimonials among the statements that need the prior-results disclaimer. Firms that publish testimonials should apply the same quoted-alone test to every testimonial that states an outcome.
- 1.Cornell LII, 22 NYCRR 1200.7.1 (New York Rule 7.1, Advertising)
- 2.American Bar Association, Comment on Model Rule 7.1
- 3.The Florida Bar, Rules Regulating The Florida Bar (Chapter 4, Rule 4-7.13)
- 4.Texas Center for Legal Ethics, Communications concerning a lawyer's services (Rule 7.01)
- 5.Harvard JOLT Digest, AI as the new front door to legal services (Dean Cook, January 5, 2026)
- 6.Pew Research Center, Google users are less likely to click on links when an AI summary appears (July 22, 2025)
- 7.EBU, News Integrity in AI Assistants (October 2025)
- 8.Press Gazette, BBC research on AI chatbot news accuracy (February 2025)