Google AI Overviews for accountants: what changed and what to fix

Google's AI Overviews now answer a large share of accounting-related searches directly on the results page — "how much does a tax accountant cost UK," "do I need an accountant for a limited company," "best small business accountant near me" — before a searcher ever reaches a blue link. If your firm's site isn't structured to be the source Google pulls from, you don't lose a ranking position. You lose the click entirely.

This is a different game from classic SEO, and most accounting-firm websites are built for the old one. Here's what changed, how to check where you stand, and a 14-day list to fix the highest-leverage gaps.

How is AI Overviews different from classic accounting-firm SEO?

AI Overviews synthesizes an answer from multiple sources and shows it above the organic results, so ranking #1 no longer guarantees visibility — being the source the synthesis cites does. Classic SEO rewarded pages that matched a keyword closely and built enough backlinks to outrank competitors for that exact phrase. AI Overviews works differently: Google's generative layer pulls fragments from several pages, blends them into a direct answer, and links out to (usually) three to five sources underneath it.

For accounting firms this changes what "ranking" means in three concrete ways.

First, long-tail, question-shaped queries now matter more than head terms. Nobody used to build a page around "can I claim mileage if I work from home UK 2026" — that traffic was assumed to funnel through a generic "tax deductions" page. AI Overviews rewards the firm that answers the specific question in a specific, extractable block, not the firm with the broadest page on the topic.

Second, corroboration across sources counts more than a single authoritative page. If your firm's answer on VAT registration thresholds matches HMRC's published figure and phrasing pattern, and three other reputable sites say something similar, Google's overview is more likely to cite you as one of the supporting sources. A page that's technically correct but phrased in a way that contradicts or oddly reworded the primary source (HMRC, IRS, Companies House) is easier for the model to skip.

Third, local intent gets resolved differently. "Accountant near me for a sole trader" used to trigger the Local Pack plus organic results. Now it often triggers an AI Overview that names specific practice types and cost ranges before the map pack even appears. If your location and service-area signals are thin, you're invisible in the part of the page that's actually getting read.

None of this means keyword density or backlink count stopped mattering — they didn't. It means they're no longer sufficient. The new requirement layered on top is *extractability*: can a model lift a clean, self-contained, factually corroborated answer from your page in one or two sentences?

Why do answer-ready pages, local proof, and corroboration matter more than keyword stuffing?

They matter more because AI Overviews rewards pages built to be *lifted*, not pages built to be *found*. Keyword stuffing helped a page get found by matching search terms. It does nothing to help a language model extract a usable answer, and it often actively hurts — dense, keyword-repetitive copy is harder to summarize cleanly than a direct, well-structured answer.

Take a concrete example. A firm's "personal tax return service" page that opens with three paragraphs of generic copy about "navigating the complexities of self-assessment in today's regulatory landscape" gives the model nothing to extract. A page that opens with "Self-assessment tax returns for sole traders and landlords, from £180 + VAT, filed within 10 working days of receiving your records" gives the model a complete, citable unit: service, audience, price, turnaround.

Local proof matters for the same reason. A firm serving Leeds and the surrounding area that never mentions Leeds, Wakefield, or Bradford by name outside a footer address is asking Google to infer local relevance from a Google Business Profile alone. A firm that names the specific towns and industries it serves — "we act for hospitality and construction businesses across West Yorkshire" — gives the model corroborating local signal that matches how people actually phrase local queries.

Corroboration matters because AI Overviews is explicitly designed to reduce the risk of citing a single, unverified source for factual claims. When your VAT threshold page states the current registration threshold, phrase it consistently with HMRC's own wording and cite HMRC directly. When your US-facing content references IRS filing deadlines, cite the IRS. This isn't just good practice — it's the mechanism by which the model decides your page is safe to quote.

What technical checks should an accounting firm run first?

Run these five checks before touching any content — technical blockers make good content invisible regardless of how well it's written.

Crawlability and rendering. Confirm Google can actually reach and render your service pages. Use Google Search Console's URL Inspection tool on your five highest-value pages (personal tax, company accounts, payroll, bookkeeping, and your local landing page if you have one). If any come back "Discovered — currently not indexed" or "Crawled — currently not indexed," that page cannot appear in an AI Overview citation no matter how well-written it is. Check your robots.txt hasn't accidentally disallowed a service subfolder — this happens more often than firms expect after a website migration.

Schema markup. Add structured data that tells Google explicitly what kind of entity and content it's looking at. For an accounting firm, the priority schema types are: - LocalBusiness or AccountingService on your homepage and location pages, with name, address, telephone, priceRange, and areaServed populated. - Person schema for each named accountant or partner, linked to their bio page, with jobTitle and professional body membership (ACCA, ICAEW, AICPA, CPA) included as hasCredential or in the description. - FAQPage schema on any page with a genuine Q&A section — this is one of the more direct routes into AI Overview citations because the format already matches how the overview presents answers. - Service schema per offering (tax returns, payroll, VAT, bookkeeping) with name, description, and where possible offers with a price or price range.

Validate all of it in Google's Rich Results Test before assuming it's working — malformed JSON-LD is worse than none, since it can trigger warnings in Search Console without you noticing for months.

Author and entity markup. Every technical article or advice page should be attributed to a named, credentialed person, not "Admin" or the firm name alone. Link the author name to a bio page that states their qualification (ACCA, ACA, CPA, EA), the firm they work for, and ideally a professional register link (ICAEW "Find a Chartered Accountant," ACCA's directory, or your state board listing in the US). Google's quality guidelines describe this as part of assessing expertise and trustworthiness, and AI systems trained partly on the same signals treat named, verifiable authorship as a meaningful corroboration signal — an anonymous "the team at XYZ Accountants" byline gives a model nothing to verify.

FAQ visibility. Check whether your FAQ content is actually rendered in the page's HTML or hidden behind a JavaScript accordion that requires a click to reveal. Many site builders load FAQ answers only on interaction, which means the crawler may see an empty <div>. View your page's rendered source (right-click → "Inspect," check the Elements tab) and confirm the full answer text is present without a click. If it isn't, that FAQ content may as well not exist to a crawler.

Internal links. Audit whether your blog content links to the relevant service page using descriptive anchor text, and whether that service page links back to supporting proof (case studies, pricing, FAQ). A post titled "Do I need to register for VAT?" that never links to your VAT registration service page is missing an obvious opportunity — both for user navigation and for reinforcing to Google which page is the authoritative destination for that topic.

What content checks matter beyond the technical layer?

Technical fixes make your content eligible to be crawled and parsed; content fixes make it worth citing. Run these checks page by page.

Niche proof. State plainly and specifically who you serve. "We help small businesses" is not niche proof. "We act for 40+ SaaS and e-commerce businesses across the UK, including VAT MOSS and Xero-to-Amazon reconciliation" is. If your firm has a genuine specialism — contractors, dentists, hospitality, crypto traders, US expats — that specificity is exactly the kind of detail that differentiates your page from a hundred generic competitors and gives the model a reason to prefer your answer for that niche query.

Location proof. Name your service area explicitly and repeatedly across relevant pages — not just in the footer. If you have multiple offices, give each one its own page with a distinct address, phone number, and locally-relevant content (mentioning the local Chamber of Commerce, local business rates quirks, or regional grant schemes you've helped clients access). A single "Locations" page listing three addresses in a table is weaker than three individual pages each answering "accountant in [town]" on its own terms.

Pricing and process clarity. State a price or a clear price range wherever you can, and describe the actual process step by step. "Contact us for a quote" is the single most common reason a service page fails to get cited — it gives the model nothing concrete to extract. Compare "Limited company accounts from £75/month, includes bookkeeping software, year-end accounts, and Companies House filing" against "Every business is different, so get in touch for a tailored quote." The first is a citable fact. The second is a dead end.

Review language. Audit what your testimonials actually say, not just that you have them. A five-star review that says "Great service, very happy!" contributes nothing extractable. A review that says "Switched from our previous accountant who missed our VAT deadline twice — this firm caught an error in our prior year's return and got us a £2,000 refund" contains specific, corroborating detail a model can use to support a claim about competence or responsiveness. Where possible, ask clients for reviews that mention the specific service used and a concrete outcome, and mark review content up with Review and AggregateRating schema so it's machine-readable, not just human-readable.

Direct-answer sections. Every service and advice page should open with a two-to-three sentence block that fully answers the page's core question before any further elaboration. This is the single highest-leverage content change available to a small firm. If the page is about self-assessment penalties, the first paragraph should state the penalty amounts and deadlines plainly, sourced from HMRC, before the page moves into detail, caveats, or a call to action. Bury this in paragraph four and you've made it much less likely to be the block an AI system lifts.

What's the 14-day action list for a small firm with limited marketing time?

Treat this as a fixed sprint, not an ongoing project — most of it is one-time structural work, not content production you'll repeat monthly.

Days 1–2: Audit your five priority pages. Pick your homepage, your two highest-revenue service pages (typically personal tax and company accounts, or payroll and bookkeeping), your local landing page, and your FAQ or advice hub. Run Search Console's URL Inspection on each. Note anything not indexed.

Days 3–4: Fix crawlability blockers. Resolve any indexing issues found on days 1–2 — check robots.txt, check for accidental noindex tags left over from a staging site, submit an updated sitemap in Search Console.

Days 5–6: Add or repair schema markup. Implement LocalBusiness/AccountingService, Person, and Service schema on the five priority pages. Validate each with Google's Rich Results Test. If you use WordPress, a plugin like Schema Pro or RankMath can generate most of this without a developer; if you're on a custom build, budget a few hours of developer time.

Day 7: Rewrite your homepage opening paragraph. Replace generic framing with a direct-answer block: who you serve, where, and the core services, in plain language, in the first 40–60 words.

Days 8–9: Rewrite your top two service pages for extractability. Add a two-sentence direct answer at the top of each. State a price or price range. State the process in three to five numbered steps. Remove vague CTAs like "contact us to discuss" as the *only* pricing information — keep the CTA, but pair it with a real number.

Day 10: Build or repair your FAQ section. Write five to eight genuine questions clients actually ask (pull these from your inbox or your intake calls — don't invent them). Answer each in two to four sentences, cite HMRC, IRS, or Companies House by name where the answer touches a factual threshold or deadline, and mark the section up with FAQPage schema. Confirm the answers render in raw HTML, not behind a JavaScript toggle.

Day 11: Add named authorship to your advice content. Attribute your three most-visited blog posts to a real, credentialed person with a linked bio page. Add their professional body membership.

Day 12: Fix internal linking. Go through your last 10 blog posts and add a direct link to the relevant service page using descriptive anchor text (not "click here").

Day 13: Audit and upgrade your reviews. Pull your five most recent reviews. For any that are vague, reach out to those clients and ask if they'd add a sentence of detail, or replace them with newer, more specific ones. Add Review schema if your site platform supports it.

Day 14: Re-run URL Inspection and Rich Results Test on all five priority pages. Confirm indexing status, confirm schema validates cleanly, and log today's date so you can compare Search Console impressions and click data in 4–6 weeks — that's roughly the window in which Google's indexing and any AI Overview citation patterns should show the effect of the changes.

None of this guarantees a citation in any specific AI Overview — Google doesn't publish selection criteria, and citation behavior varies by query and shifts as the system is retrained. What this sprint does is remove the structural reasons your firm's genuinely useful content is currently ineligible to be considered at all. For a firm with one or two people handling marketing alongside client work, that's the realistic, finishable goal for two weeks: not guaranteed visibility, but removed exclusion.