You manage three mid-market CPA firm clients. Tax season just wrapped, they each closed between eight and twenty new engagements, and not one of those new clients can tell your partner-in-charge exactly how they found the firm. Referrals still carry the practice — 58% of accounting firm clients still arrive through referrals — but the other 42% are landing through channels you cannot see on the GA4 dashboard. Increasingly, that invisible channel is an AI engine.
Try the test yourself this afternoon. Open ChatGPT and ask “recommend a CPA for a venture-backed startup in Austin.” Open Perplexity and ask “best accounting firm for SaaS companies under $10M ARR.” Your client almost certainly does not appear. 77% of businesses ranking on Google page 1 are completely invisible in ChatGPT, and accounting is one of the lowest-visibility verticals. The reason is mechanical: accounting queries pull heavily from trusted mega-domains. NerdWallet, Bankrate, TurboTax, H&R Block, Investopedia, and IRS.gov dominate AI-generated answers for tax and accounting questions. Individual CPA firms get crowded out before the prospect ever sees a name to call.
The cost compounds. 94% of B2B buyers now use LLMs for research, and they purchase from their Day One shortlist 95% of the time. If your client is not on the shortlist the AI engine surfaces, no follow-up automation pulls the deal back. This playbook is the working document we hand to agency teams running their first AEO retainer on accounting clients — four phases over 90 days, with opinionated calls on what to skip.
Run the playbook on your own client list.
Growth plan free for 14 days. Five AI engines. Full agency dashboard.
Start free trialThe trusted-domain trap that keeps CPAs out of AI answers
Start with the shape of the market. 50,885 CPA firms operate in the US, employing 572,995 people and generating roughly $151 billion in annual revenue. The U.S. Accounting Services industry market size is $157.4 billion this year, with a 5-year CAGR of 1.3%. A massive, low-growth, high-competition market — which is exactly the configuration where AEO visibility moves the needle most.
The dominance of trusted mega-domains is the core bottleneck. When a prospect asks ChatGPT about quarterly estimated payments, R&D tax credits, ERTC eligibility, or multi-state nexus, the engine cites the sources it has been trained to treat as authoritative. NerdWallet, Bankrate, TurboTax, H&R Block, Investopedia, and IRS.gov dominate AI-generated answers for tax and accounting queries. These are the domains that appear in every “best tax guide” roundup, hold the most incoming links, and live at the top of Google for years. AI engines inherit that ranking signal.
Individual CPA visibility suffers as a consequence. The solo CPA who has written the most lucid guide to founder tax strategy on the internet does not show up unless her brand name has earned mentions in publications the engines already trust. The agency playbook changes — stop optimizing for ranking signals, start optimizing for citation signals. Brand mentions correlate 0.664 with AI visibility; backlinks correlate 0.218. Roughly a 3-to-1 advantage for mentions over links.
Do not promise your CPA client “top of ChatGPT for best CPA in [city].” Broad geographic queries are owned by review aggregators and large national firms for years to come. The winnable lane is niche — ERTC, R&D credits, multi-state nexus, dental-practice CPA, SaaS founder taxes. Set that expectation in the discovery call, in writing, before the retainer signs.
The compliance layer most AEO agencies skip
Before any phase work begins, the guardrails. Accounting is regulated, and AI-driven content needs an extra review layer that most generalist agencies are not used to running.
The AICPA Code of Professional Conduct (Rule 502) prohibits advertising that is false, misleading, or deceptive; advertising claims must not create unjustified expectations or state fixed fees that are not actually guaranteed. That language matters in practice. Your CPA client cannot claim “we will save you X dollars in taxes” or “guaranteed lowest tax bill” in any AEO content unless the claim is documented, conditional, and demonstrable for the segment described.
Circular 230 (the IRS regulation governing tax practice) prohibits any public communication containing false, fraudulent, coercive, misleading, or deceptive claims; enrolled agents must use permitted terminology and may not imply an employer relationship with the IRS. If the firm employs enrolled agents or unenrolled preparers, terminology in the FAQ schema and bio pages has to match. No use of “certified” outside the credential, no IRS seals, no copy that implies an IRS partnership.
The practical implication is workflow. Every piece of AEO content runs through a compliance filter before publication. Soft language by default: “typically save,” “may help,” “depending on your situation.” Every tax claim ties back to a source or a caveat. Sell compliance review as part of the retainer, not as friction — agencies that handle this close CPA retainers faster than generalists who get caught flat-footed in the partners’ review.
The 90-day playbook in four phases
Four phases, each scoped to a realistic 21-to-25 day window with overlap at the seams. The phases run in order but you can begin Phase 2 schema work in parallel with Phase 1 baseline measurement — the engineering pace decouples cleanly from the measurement pace. The deliverable for each phase is named in the eyebrow, and the bullets below it are the working checklist your team executes against.
Audit, baseline, and define the winnable lane (Days 1–21)
Most CPA firms have never seen their own AEO numbers. The audit is the first artifact that anchors the retainer and gives your team the data needed to write a defensible Q1 plan. Move quickly — this phase should not consume more than three weeks.
- ▸ Run 10–20 shopper-intent queries across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. Capture per-engine mention rate, position of first mention, and total mentions.
- ▸ Score the baseline using a published methodology so the number is defensible. The GenPicked AEO Citation Score weights ChatGPT 0.35, Perplexity 0.25, Gemini 0.25, Claude 0.15 and dropping any failed engine.
- ▸ Screenshot the top five queries on every engine on the same calendar day. Store the screenshots in the client folder — they become the “before” image in the monthly report.
- ▸ Interview the firm partner. Ask which three niches generate the highest-margin engagements and which queries those clients used during discovery.
- ▸ Write a one-page baseline brief: current ACS band (invisible / emerging / competitive / category-leader), three named niches, ten target queries per niche, the trusted-domain competitor set the firm cannot beat.
Schema stack and content chunking (Days 14–42)
The engineering work overlaps with Phase 1 by a week so your developer is not sitting idle. Schema is foundational — without it, the content in Phase 3 underperforms by a measurable margin.
- ▸ Implement LocalBusiness schema with full address, hours, phone, and aggregateRating where review counts allow.
- ▸ Layer AccountingService schema referencing the named service types: tax preparation, audit, bookkeeping, advisory, tax planning.
- ▸ Build FAQPage schema with 10–15 Q&A pairs per niche, each answer written as a self-contained 80–140 word chunk. Pages with FAQPage schema are 3.2× more likely to appear in Google AI Overviews.
- ▸ Validate every schema block in Google’s Rich Results Test and Schema Markup Validator before deploying. Broken schema is worse than missing schema.
- ▸ Verify each chunk reads as a stand-alone answer. If an AI engine atomizes the FAQ into individual snippets, the snippet still has to make sense on its own.
Niche content build-out (Days 28–63)
This is where the firm earns the right to be cited. Pick three to five niches in Phase 1, then create the destination pages an AI engine will land on when an actual prospect asks a niche-specific question. Generic “FAQs about our firm” pages do nothing.
- ▸ Write one pillar page per niche: 1,800–2,400 words, 12–18 H2/H3 subheads, FAQ block at the bottom, internal links to relevant service pages.
- ▸ Use real client scenarios (anonymized) rather than abstract examples. AI engines reward specificity — a worked example beats a definition every time.
- ▸ Run every claim through the compliance filter. Replace “will save” with “may help reduce,” cite the IRC section for every quantitative claim, document the worked-example assumptions.
- ▸ Build a fifteen-question FAQ block per niche, formatted as 100–150 word chunks with Q&A headings. For niches like R&D credit consulting, AI can analyze project management tickets to identify qualifying work, calculate exact percentages, and flag duplicates — lean into that operational detail in the FAQ.
- ▸ Ship pillar pages in series, not parallel. One pillar live, indexed, and earning mentions before the next one drafts.
Earned mentions and continuous monitoring (Days 35–90)
The slow, high-impact work. Earned mentions on publications AI engines trust are the single biggest visibility lever, and they cannot be shortcut. Start the pitch sprint as soon as the first pillar page is live so the journalist has something to link.
- ▸ Build a target list of 25 outlets: Journal of Accountancy, Accounting Today, CPA Practice Advisor, Bloomberg Tax, local business journals, three relevant subreddits, two specialist newsletters.
- ▸ Pitch the firm partner as a source on a single named niche. One specialty per pitch — the prospect that says yes will rarely buy a generalist.
- ▸ Track every mention in a spreadsheet: outlet, link, date, the exact sentence the firm is named in. The sentence is what AI engines parse.
- ▸ Run daily citation sweeps across all five engines. Build the monthly client report around the diff: new mentions, lost mentions, position changes, competitor mentions gained or lost.
- ▸ Deliver the monthly report by day 15 of each month, white-labeled if the agency plan supports it. Lead with share of voice against the local and national competitor set, not raw scores.
Do not stack all four phases into the first thirty days. The compression breaks two things: the compliance review timeline (partners need a real two-week turn) and the earned-mention pipeline (journalists do not respond to pitches with no published proof point). Phases overlap on purpose, not by accident.
Pricing the retainer using the 12–15% allocation anchor
The budget math is cleaner than most agency conversations make it. Enterprise CMOs allocated an average of 12% of their digital marketing budget to AEO/GEO last year; competitive organizations are pushing 15% or above this year. 94% of CMOs are planning to increase AEO/GEO investment, and AEO/GEO ranked as the #1 strategic marketing priority for the year. 97% of digital leaders reported positive AEO/GEO impact on their organizations.
For a mid-market accounting firm with a $50,000 annual marketing budget, 12–15% = $6,000 to $7,500 per year, or $500 to $625 per month. A defensible monthly stack: GenPicked Growth plan ($197/month) plus one per-brand Lite tier ($75/month) = $272/month, comfortably under the ceiling with room for service hours. When the firm layers in a second niche, upgrade to Growth plus Standard ($149/brand) = $346/month baseline, still inside the 12% benchmark.
Anchor on the percentage in the proposal, name the GenPicked line items, then build service hours on top. Partners read the number as a reallocation rather than an increase, and the conversation moves to execution faster.
Schema specificity beats schema volume
Most AEO advice for accounting firms stops at “add FAQ schema.” The advice is incomplete and occasionally counterproductive. Sites with proper schema are cited by Perplexity 67% more often and appear in ChatGPT responses 3.2× more frequently, but the lift only shows up when the schema is specific to the entity it describes.
Generic Article or Organization schema on a CPA firm page can actually underperform no schema at all because it sends a noisy signal. The stack that works: LocalBusiness for the physical office, AccountingService for named service types, FAQPage for niche Q&A blocks, Person schema for each partner with credential properties matching the licensure record. Each block points to the next via @id references so the engine reads a connected entity graph.
The chunking inside each FAQ matters as much as the schema wrapping. 78% of AI-generated answers include list formats; FAQ schema naturally structures content as Q&A pairs, matching the exact format AI platforms present to users. The firm that owns niche FAQ pages on ERTC eligibility, home office deductions for self-employed CPAs, multi-state nexus thresholds, and equity compensation timing will pull citations the firm with a generic FAQ page never sees.
Niche verticalization is the defensible positioning
The pattern that keeps showing up: the firm wants to position as general-purpose accountants because that matches how the practice operates. The instinct is correct for the business model and wrong for AI search. Broad queries (“best accountant near me”) are owned by review aggregators and large national firms. Niche queries are wide open.
Pick three to five niches per firm: R&D tax credits for tech founders, ERTC eligibility work, multi-state nexus for e-commerce sellers, dental-practice accounting, SaaS founder taxes and equity compensation for early-stage operators. Each niche gets a dedicated cluster: pillar page, FAQ block, two or three supporting posts, schema graph tied together. A mid-market firm that claims authority over three niches in the first 90 days will land in the emerging band (20–39 ACS), and that is category-leader positioning inside the niche even if the firm is invisible on the broad query. The niche play also unlocks AI Overview citations from page 2 or page 3 organic positions — Ahrefs documented that AI Overviews cite sources well below the page-1 ceiling when entity match and content structure justify it. The competitive moat is unusually narrow because most accounting firms have zero AEO surface area — the window is measured in months, not years.
The attribution gap your GA4 dashboard is hiding
When an AI engine sends a prospect to your client’s site, the conversion usually shows up in GA4 as “direct.” ChatGPT, Perplexity, and several other engines do not reliably pass referrer headers. The baseline expectation: most of your AEO-driven traffic is invisible inside the analytics tool you use to prove the retainer.
The fix is partial but essential. Build custom GA4 channel groups that identify AI-sourced traffic by user-agent patterns, niche FAQ landing-page entries, and disciplined UTM tagging on every earned-mention link. Once corrected, the data shifts visibly: AI visitors who do click through spend 68% more time on the site than organic search visitors and convert at higher rates. The traffic was always there.
Reporting that defends the retainer
The monthly report is the artifact that determines retainer survival. Build it around three numbers: share of voice against the named competitor set, new mentions earned in the period, and lost mentions flagged for risk. Lead with the ACS band rather than the raw score to keep the conversation strategic. Deliver on the same calendar day each month so partners can budget the review meeting.
Include the screenshot diff from the audit phase. The before-and-after panel of the same query on the same engine, six months apart, is the most persuasive artifact you can put in front of a CPA partner. That picture closes renewals.
Run the playbook on your own client list.
Growth plan free for 14 days. Five AI engines. Full agency dashboard.
Start free trial