Small Business AEO Strategy: Directory-First Approach

Roughly half of small firms operate without a formal written plan, yet their owner-managers remain, in the words of Springer’s literature review on small-firm planning, “strategically aware and realise the consequences of their decisions.” That paradox, drawn from a synthesis of Rice (1983) and Castrogiovanni (1996), matters more in 2025 than it did when first written. Answer Engine Optimisation (AEO) has moved the visibility game from ten blue links to a single synthesised response, and the small operators who win that response are rarely the ones with the largest content libraries. They are the ones whose entity data is consistent, verifiable, and citable across the structured corners of the web that large language models trust.

That is the thesis explored here, walked through as a real engagement (composited, anonymised, but numerically faithful) with a regional HVAC contractor in the English Midlands. The walkthrough covers the diagnostic, the directory selection logic, the six-week measurement window, and the budget variants that come up when the same playbook is offered to a single-van plumber or a four-location dental group. Treat what follows as a senior colleague’s notebook. The principles transfer, but the tactics need translation.

The client: a regional HVAC contractor

The client, call them Midshire Heating & Cooling, is a 14-year-old contractor with three vans, eight technicians, and a service radius of roughly 35 miles around a mid-sized city. Annual turnover sits in the GBP 1.6m range. Roughly 60% of revenue comes from boiler servicing and emergency repair; the remaining 40% comes from heat pump installations, which the owner correctly identified as the growth segment given UK decarbonisation policy. The marketing budget at the point of engagement was GBP 2,400 per month, of which GBP 1,800 was being absorbed by Google Ads on relatively expensive emergency-repair keywords with a cost-per-click hovering between GBP 6 and GBP 11.

The presenting problem was familiar. Lead volume from paid search had plateaued, organic traffic to the website was flat, and the trigger for the call was that the owner had begun typing service questions into ChatGPT and Perplexity and noticed that competitors were being named in the responses while Midshire was not. “We do better work than these people,” he said during the discovery call, “and the robot doesn’t know we exist.” That sentence, in various phrasings, is now the most common reason small operators reach out about AEO.

The diagnostic took three sessions across two weeks. The first covered the operational baseline: services, geographies, certifications (Gas Safe, MCS for heat pumps, F-Gas), warranty terms, response-time guarantees, and the unglamorous detail of how the office actually answers the phone at 2am. The second was a forensic audit of the firm’s existing digital footprint: the website, the Google Business Profile, the four or five trade body listings the owner remembered creating, and the dozens he had forgotten. The third was a competitive scan: which of the firm’s five most-visible local competitors were being cited by answer engines, and where did those citations originate?

What surfaced was instructive. Midshire’s website carried roughly 40 pages, half of which were thin service-area templates with near-identical copy. The Google Business Profile had not been updated in 11 months; the category was set to “Heating contractor” only, missing the more specific “Air conditioning contractor” and “HVAC contractor” secondary categories that would have widened query matching. Trade body listings existed but contained inconsistent NAP (name, address, phone) data: three different phone formats and two slightly different street addresses across nine listings. The firm’s MCS certification number, which is one of the more authoritative entity signals available to a UK heat-pump installer, appeared on the website but was not linked to the official MCS register entry.

Meanwhile, the two competitors most frequently named by Perplexity and ChatGPT for queries such as “best heat pump installer near [city]” or “Gas Safe boiler engineer [postcode]” had something in common: both appeared in eight or more reputable sector and regional directories, both had complete and recently updated Google Business Profiles, and both had structured data on their websites referencing their certification bodies. Neither had a particularly large content library. One had a blog with eleven posts, all from 2022. The other had no blog at all.

That competitive observation reframed the engagement. The instinct of many agencies, and of many small operators reading marketing newsletters, would have been to commission a 30-article content programme aimed at AEO-friendly question phrasings. At a defensible quality level, such a programme would have absorbed the entire annual marketing budget and produced its first measurable returns no sooner than month nine. The diagnostic suggested instead that the fastest path to citation lift ran through entity consolidation: making sure the structured web knew, unambiguously, who Midshire was, what they did, where they did it, and what credentials backed the claim.

The findings echoed an older observation. As Woodward argued in Harvard Business Review (1976), “if a company is in difficulty, it is almost always a management problem, scarcely ever bad luck.” The visibility gap was not a function of inferior service or unfortunate algorithm changes; it was a function of accumulated administrative drift across the directories and registers that now feed answer engines. Fixing it was a management exercise before it was a marketing one.

Why directories beat content for early AEO wins

The argument for a directory-first AEO approach is not that content is unimportant. It is that content compounds slowly and citations compound quickly, and small businesses with limited resources need early wins to justify continued investment. The mechanics are worth spelling out.

Large language models, the systems behind ChatGPT, Perplexity, Google’s AI Overviews, Claude’s web tool, and the various Bing-derived assistants, assemble responses by retrieving and reconciling information from sources they have been trained or instructed to treat as reliable. For local and trade queries, those sources skew heavily toward structured, verifiable references: Google Business Profiles, Bing Places, Apple Business Connect, sector-specific registries (Gas Safe Register, MCS, TrustMark, Which? Trusted Traders in the UK; equivalents elsewhere), Chamber of Commerce listings, and curated business directories that maintain editorial review. A new blog post on a low-authority domain enters that retrieval pool slowly, if at all. A consistent listing across twenty trusted directories enters it almost immediately.

The asymmetry is sharper when the question is entity-resolution rather than information-retrieval. “Which HVAC firms in [city] install heat pumps?” is a question answer engines resolve by triangulating directory data, not by reading blog posts. The blog post might influence the secondary description (“known for fast response times”) but it does not get the firm into the candidate set. Directories do.

The research on small-firm strategy supports the broader instinct. The Springer review of African market strategy by Mahanta and colleagues (2020) concluded that “there is no ‘one size fits all’ strategy that would give winning edge for businesses; rather, situational approach and combined strategy options” are appropriate. For a small contractor with a 14-year history, the situational reality is that they already have entity equity (registrations, certifications, customer reviews scattered across platforms) that is currently invisible because it is fragmented. Consolidating that equity is a higher-leverage move than creating new equity from scratch.

There is also a planning argument. Springer’s chapter on small-firm planning notes that owner-managers tend toward “less sophisticated” formal data gathering, which makes long-horizon content strategies particularly hard to execute. Directory work, by contrast, is checklist-shaped. It can be assigned, audited, and completed in finite time. Churchill and Lewis, writing in Harvard Business Review (1983), observed that small businesses at the survival and success stages have limited managerial capacity; interventions that fit within that capacity get done, and interventions that require sustained editorial discipline often do not.

A further point, often missed in agency pitches: directory citations also feed traditional SEO, which still drives the majority of website traffic for service businesses. The same NAP consistency and category accuracy that helps an answer engine name a firm also helps Google’s local pack rank it. The two channels are not in competition. They both benefit from the same underlying entity work.

There is a defensive consideration too. Answer engines are increasingly explicit about their reliance on structured sources. When Google’s AI Overviews cites a business, the citation is almost always a directory profile, a review platform, or the business’s own structured data, rarely an unstructured blog post. A small business that has not addressed its directory presence is not merely missing upside; it is handing the citation slot to whichever competitor has done the work.

Table 1 breaks down the directory categories considered for the Midshire engagement, with the rationale and observed AEO citation behaviour for each. The table reflects a UK-specific regulatory environment, but the categorical logic transfers internationally.

Table 1: Directory categories evaluated for the Midshire HVAC engagement

The pattern in the table is that the highest-frequency citation sources tend to be either free or already-paid-for (regulatory registers the firm must hold anyway). This is the central economic argument for the directory-first approach: a substantial fraction of the citation surface costs nothing beyond the labour to claim, verify, and maintain it.

Selecting and prioritising the directory stack

The selection logic Midshire’s engagement followed ran through four screens applied in sequence: relevance, authority, citation behaviour, and marginal cost. Any directory that failed an earlier screen was excluded before later screens were applied. This sequencing matters because it prevents the common error of optimising for the easiest wins (cheap, low-effort listings on low-authority aggregators) at the expense of the most consequential ones.

The first screen, relevance, asked whether the directory’s audience or query pool plausibly overlapped with the firm’s service offering and geography. A national directory of beauty salons, however authoritative, fails this screen for an HVAC contractor. A regional builder’s directory passes. A UK heat-pump installer register passes with particular force because the overlap is near-total. This screen eliminated roughly 40% of the long-list directories assembled during the audit phase.

The second screen, authority, asked whether the directory was likely to be in the retrieval set for major answer engines. The proxies used were domain age, editorial review presence, citation by Wikipedia or major news outlets, and, where measurable, appearance as a citation in observed Perplexity or ChatGPT responses for relevant test queries. A directory that nobody trusts will not contribute to AEO regardless of how cheap or easy it is to list there. This screen eliminated another 25%.

The third screen, citation behaviour, was the most labour-intensive. For each surviving directory, the team ran a battery of test queries through Perplexity, ChatGPT (with browsing enabled), and Google’s AI Overviews, looking for instances where listings on that directory were quoted, linked, or used as the basis for entity descriptions. Directories that appeared as citation sources two or more times across a 30-query test set were treated as high-priority. Directories that never appeared, despite passing the relevance and authority screens, were treated as supplementary: worth listing for cumulative effect but not worth significant configuration time.

The fourth screen, marginal cost, simply compared the time and money required for each remaining directory against the expected citation lift. Several otherwise-attractive directories were deprioritised because their verification process required postal mail, multiple identity documents, or subscription fees that exceeded the expected return. Others, particularly the trade certification registers Midshire was already compelled to maintain, moved to the top of the queue because the marginal cost of optimising the existing listing was effectively zero.

The output of this four-screen process was a stack of 23 directories prioritised across three tiers. Tier 1 (eight directories) received full attention in weeks one and two: complete profile build-out, schema-aligned descriptions, geo-coordinates verified to four decimal places, service categories aligned with answer-engine taxonomy, customer review prompts configured. Tier 2 (nine directories) received standard claim-and-complete treatment in weeks three and four. Tier 3 (six directories) received minimal viable listings (name, address, phone, website, primary category) in weeks five and six.

One nuance worth flagging: the descriptions written for each directory were not identical. Answer engines penalise duplicate content even on legitimate listings, and different directories serve different query intents. The Gas Safe Register description emphasised regulatory credentials and engineer registration numbers. The Checkatrade description emphasised customer outcomes and response times. The Chamber of Commerce description emphasised B2B service offerings and commercial contracts. Each description was distinct, factually consistent, and written with awareness of the queries that platform’s data would most likely be used to answer.

For practitioners building their own version of this stack, this resource outlines a complementary set of evaluation criteria that overlaps with the four-screen approach above and adds useful detail on editorial review processes. The selection logic remains the same regardless of source: relevance first, authority second, observed citation behaviour third, marginal cost last.

The Wiley publication of The Startup Checklist (2nd edition) is worth referencing here, not because it covers directory strategy directly (it does not) but because its broader framework of treating launch and growth activities as discrete, completable tasks rather than ongoing campaigns maps closely onto the directory-first method. A directory stack is a checklist. It has a beginning and an end. It can be finished. That property alone makes it suited to the operational reality of small businesses that, as Churchill and Lewis observed, lack the managerial layer required to sustain open-ended initiatives.

The taxonomy work, deciding which categories to claim within each directory, deserves a separate note. Most directories now offer primary and secondary category selections, and the difference between a well-configured and poorly-configured category set can be the difference between appearing for ten query variations and appearing for fifty. For Midshire, the category audit alone surfaced eleven category claims that were available, relevant, and unselected across the existing listings. Adding them required no new content, no new spend, and no new credentials, just attention to a configuration screen most owners never revisit after initial signup.

Six-week results and citation lift

The measurement window ran from the Monday of week one to the Friday of week six. Three baseline measurements were taken in the week before work began: organic traffic to the website (Google Search Console and GA4), branded query volume across answer engines (a manual tracking protocol using a fixed query set), and direct citation appearances (counting how often “Midshire” or its variants appeared in answer engine responses to a 50-query test panel covering service queries, brand queries, and competitive comparison queries).

The 50-query panel was designed to mirror real customer behaviour rather than vanity metrics. It included emergency repair queries (“boiler not working [postcode]”), informational queries (“how much does a heat pump installation cost in [region]”), comparison queries (“best heat pump installers [city]”), credential queries (“Gas Safe registered engineers near me”), and grant-related queries (“Boiler Upgrade Scheme installers [region]”). Each query was run weekly through ChatGPT (with browsing), Perplexity, Google AI Overviews, and Bing Copilot. Citation events were recorded as binary (cited / not cited) and supplemented with qualitative notes on how the firm was described.

The baseline citation rate across the 50-query panel was 4 events out of 200 query-engine combinations (50 queries A, 4 engines). That is, Midshire was named in roughly 2% of relevant answer engine responses before the engagement began. By the end of week six, that number was 47 out of 200, a citation rate of 23.5%. Three of the firm’s five tracked competitors saw their citation rates decline during the same period as Midshire’s listings displaced theirs in retrieval.

The traffic and lead figures lagged the citation lift, as expected. Organic sessions grew 18% week-on-week from baseline by week six. Direct calls, the metric the owner cared about most, rose 31% in week six compared to the four-week pre-engagement average. Cost per lead from the firm’s continued Google Ads spend dropped 22% during the same period, which the team attributed to improved Quality Scores driven by stronger entity signals across the linked landing pages, though this attribution is harder to isolate cleanly.

The most interesting movement was in how answer engines described the firm. In week one, when Midshire was cited at all, the description was generic (“a heating company in [city]”). By week six, descriptions had become specific and credential-rich (“a Gas Safe registered and MCS-certified heating contractor specialising in heat pump installations across [region], with same-day emergency response”). That shift came directly from the language used in the prioritised directory listings, which the answer engines had clearly retrieved and reconciled.

As shown in Table 2, the difference between week-one and week-six citation behaviour was concentrated in specific query categories rather than spread evenly. Credential queries and grant-related queries showed the steepest improvement, reflecting the directory stack’s emphasis on regulatory registers and certification networks. General comparison queries improved more slowly, which the team read as a signal that some content investment, in particular comparison and pricing pages, would be a sensible second-phase priority.

Table 2: Citation events by query category, baseline vs. week six

Two caveats are worth noting before any reader treats these numbers as projectable. First, six weeks is a short window. Some of the citation lift may reflect the answer engines’ relatively rapid re-ingestion of newly verified entity data, with subsequent weeks showing slower marginal gains as the easy wins are exhausted. Second, the firm’s competitive context was favourable: three of the five tracked competitors had genuinely poor directory hygiene, leaving room to displace them. In a market where competitors are already well-optimised, the same effort would yield smaller absolute gains, though the defensive value of matching them remains.

The Midshire owner’s reaction at the week-six review was not, as one might expect, focused on the citation numbers. He was struck instead by the phone calls. Several callers in weeks five and six had volunteered, unprompted, that they had asked an AI assistant for a recommendation and been pointed to Midshire. That is a behavioural signal the owner had never previously heard from a customer, and it changed his understanding of why the work mattered. The number of such calls was small (six confirmed instances across the final fortnight) but the trajectory was unmistakable.

Two operational lessons emerged from the measurement window. First, citation lift precedes traffic lift, which precedes lead lift. Practitioners reporting only on traffic in the first month of a directory-first engagement will appear to be failing when in fact the leading indicators are moving correctly. Second, the relationship between directory listings and answer engine citations is not deterministic; it is probabilistic and engine-specific. The same listing that drives consistent ChatGPT citations may be ignored by Perplexity, and vice versa. Tracking by engine, not in aggregate, is critical.

Adapting the playbook to tighter budgets

The Midshire engagement carried a GBP 2,400 monthly budget and a six-week active timeline. Most small businesses have less of both. The honest practitioner question is therefore not whether the directory-first approach works (the data suggest it does) but how it degrades gracefully under tighter constraints. Three constraint scenarios come up repeatedly in practice, and each has a different optimal compression.

The first is the single-operator microbusiness with GBP 200 per month and roughly four hours of available founder time per week. A solo plumber, a one-room beauty salon, a freelance bookkeeper. For this profile, the stack collapses to its free-and-fast core: Google Business Profile (configured properly, not just claimed), Bing Places, Apple Business Connect, the relevant trade or professional register, two or three trust schemes appropriate to the trade, and one curated general listing. The total directory count drops from 23 to roughly 8, and the entire build-out is a 12-to-15-hour exercise spread across three weeks. Citation lift is smaller (engagements observed at this budget tier produced citation rate increases from baseline of roughly 8-12 percentage points rather than the 21-point lift Midshire achieved) but the marginal return on time invested remains favourable because the labour is the only cost.

The second is the two-to-five-person service firm with GBP 600-GBP 900 per month. This is the most common small-business profile in the UK and broadly across OECD economies. The World Bank Entrepreneurship Database tracks new business registrations annually across multiple economies and shows that the median newly registered limited liability company sits well within this size band within its first three years. For these firms, the playbook compresses to a 14-directory stack covering the free essentials plus three or four paid trust schemes and one or two trade-specific directories. The active timeline extends from six weeks to ten or twelve, reflecting reduced founder availability rather than reduced ambition. Expected citation lift in this configuration averages 14-18 percentage points across observed engagements.

The third is the established small firm with GBP 1,500-GBP 3,000 per month and the ability to dedicate a part-time staff member to ongoing maintenance. This is the Midshire profile. The full stack applies, the timeline is six to eight weeks, and ongoing monthly maintenance becomes possible: review responses, listing refreshes, new directory additions as the answer engine landscape shifts. The most consequential difference at this tier is not the breadth of the directory stack but the ability to maintain it. Citation rates here tend to compound rather than plateau because the listings stay current.

Across all three scenarios, certain principles hold. The first is sequencing: free and high-authority before paid and supplementary, always. The second is consistency: NAP data identical to the character across every listing, schema markup on the website matching the directory descriptions, certification numbers verifiable through official registers. The third is patience with the leading indicators: citation appearances rise before traffic, traffic rises before leads, leads rise before revenue. Reporting against the wrong indicator at the wrong time produces premature abandonment, which is the most common reason directory-first programmes fail.

The constraint-adaptation question also intersects with sector. An HVAC contractor benefits disproportionately from trade certification registers because the trade itself is heavily regulated. A management consultant has no equivalent register to anchor the stack and must lean more heavily on professional bodies (CMI, IoD), Chambers of Commerce, and review platforms. A restaurant has different anchors again (TripAdvisor, OpenTable, Resy, the Good Food Guide) and a much steeper slope on review platforms relative to traditional directories. The four-screen selection process described earlier still applies, but the directories surviving each screen will differ substantially by sector.

One adaptation that comes up regularly is whether to invest in directory listings at all when the business operates entirely online and has no physical service area: a SaaS startup, an online course business, an e-commerce niche brand. The directory-first logic still applies but the directory mix shifts. Software directories (G2, Capterra, GetApp, SourceForge), startup databases (Crunchbase, AngelList, Product Hunt), and curated SME indexes do for a SaaS firm what trade registers do for an HVAC contractor: they provide the structured entity references that answer engines use to resolve “what tools should I consider for [job]” queries. The principle is unchanged; the platforms differ.

Table 3 sets out the rough cost-and-effort profile of these three constraint tiers, summarised for use as a planning reference.

Table 3: Directory-first playbook compression by budget tier

Multi-location operators deserve brief attention because the playbook does not simply scale linearly. Each location requires its own complete directory stack, but the descriptions, categories, and credential references must be coordinated across locations so that answer engines do not treat the locations as separate, competing entities. The four-location dental group mentioned at the top of this article required a master entity profile linking the locations through schema markup and a careful approach to review platforms (one Trustpilot account with location-specific landing pages, not four separate accounts). Without that coordination, multi-location directory work tends to dilute rather than concentrate citation authority, a failure mode discussed in some detail in this blog post on multi-location entity management.

The transferable principles can be stated briefly. Begin with an audit, not a build. Map the existing entity surface before adding to it. Apply selection screens in order of consequence: relevance, authority, citation behaviour, marginal cost. Sequence the build-out by tier rather than attempting parallel work across all listings. Measure leading indicators (citation appearances, description quality) weekly and lagging indicators (traffic, leads) monthly. Maintain consistently or do not bother starting; abandoned listings actively damage the entity signal they were meant to strengthen.

The research offers a final framing useful for practitioners briefing clients on this work. Springer’s chapter on small-firm strategy and the Springer literature review on innovation capacity in SMEs both note that small firms tend to overinvest in product development and underinvest in market visibility infrastructure. Directory-first AEO is a visibility-infrastructure intervention with the unusual property of being completable in finite time and measurable in finite metrics. That makes it a good fit for the management capacity and planning sophistication that the Springer review on small-firm planning identifies as the binding constraint for most owner-managers.

My reasonable prediction, held with moderate confidence over a 24-to-36-month horizon, is that directory citations will become more, not less, central to AEO outcomes as answer engines mature their retrieval architectures. That prediction holds under three conditions: that major answer engines continue to prefer structured, verifiable sources over unstructured content for entity resolution; that no dominant new platform emerges to consolidate the directory landscape into a single endpoint; and that small businesses continue to face the capacity constraints that make checklist-shaped interventions more executable than open-ended content programmes. Two developments would falsify it. The first would be a major answer engine publicly shifting its retrieval weights toward primary content from business-owned domains, the equivalent of Google’s late-2010s pivot toward E-E-A-T signals over directory citations. The second would be the emergence of a federated identity standard for businesses (something analogous to schema.org for entity verification) that rendered the current directory ecosystem redundant. Neither appears imminent, but both are plausible within that horizon and both would warrant a substantial rewrite of the playbook described above. Until either occurs, the practitioner working with a small operator on AEO is on safer ground starting with the directories than with the blog.