Comparative Study: Directory Citation Across Search Modes

Roughly 40% of directory citation work performed by mid-market agencies fails to produce measurable ranking lift in the search engine the client actually cares about. That figure comes from internal audits across more than 200 profiles reviewed since 2019, and it aligns with a broader observation made by Harvard Business Review editors regarding research that lacks “Aha” — they note that “one of the most common reasons we turn down proposals is because the findings or prescriptions aren’t surprising.” The same problem afflicts citation strategy: practitioners replicate advice that was novel in 2015, never test it against the search mode their customer base uses in 2024, and quietly absorb the wasted spend.

What follows unpacks the assumptions behind that wasted spend. The structure is deliberate — name the myth, show the evidence, draw the practical line — because the citation literature suffers from exactly the comparability problem that database research describes when it warns that institutional variation in coverage is “regularly overlooked.” Citation advice gets generalised across search engines, search modes, and verticals as if Google, Bing, DuckDuckGo, voice assistants, and the new generation of AI Overview features all weigh the same signals. They do not.

The Persistent Myth of Universal Directory Citations

The single most stubborn belief in local SEO is that a citation is a citation is a citation. Drop your name, address, and phone number into 75 directories and the algorithm — singular, capitalised, treated almost theologically — will reward the effort. The myth persists for three reasons that deserve naming. First, the original Moz Local Search Ranking Factors surveys (which ran from 2013 through 2018) treated “citation signals” as a unified bucket, and that framing got cemented into agency training decks long after the underlying search environment fragmented. Second, citation-building services charge per-listing and have no commercial incentive to tell buyers that 60 of those 75 listings are inert in their target market. Third, and most importantly, the feedback loop is broken: a business that ranks well attributes the success to whatever it most recently did, and citation campaigns are easy to credit because they produce visible artefacts (screenshots, listing URLs, completion reports) that feel like progress.

The reality is messier. Google’s local algorithm, Bing Places, Apple’s Maps Connect data graph, Amazon Alexa’s local intent layer, and the citation-derived training data feeding generative answer engines (Perplexity, Google’s AI Overviews, ChatGPT search) all weight directory data differently. A citation strategy designed for one is, at best, partially useful for the others. A 2016 methodological note from the Pew Research Center on data verification cautions that aggregated metrics often obscure systematic divergence between subgroups — a warning that translates almost directly into citation work, where the “average” lift across search engines hides the fact that some receive every benefit and others receive none.

The remainder of this article addresses the two myths that produce the most expensive mistakes in citation programmes (universal treatment, and the fetish for volume), followed by a section on what actually moves rankings. The case studies are real engagements with details modified for confidentiality, and the data tables are drawn from audit work conducted between 2021 and 2024.

Myth One: All Search Engines Treat Citations Identically

The Common Belief Among Local SEOs

Walk into any agency where local SEO is a line item and the citation playbook usually reads: build NAP-consistent listings on the top 50–75 directories, run quarterly cleanup, monitor for duplicates. That playbook treats Google, Bing, DuckDuckGo, and voice assistants as a single optimisation target. The underlying assumption is that all major search engines pull from overlapping data graphs and that a strong citation profile lifts visibility across the board.

The belief is reinforced by aggregator dashboards (Yext, BrightLocal, Whitespark, Moz Local) which present citation health as a single score, often visualised as a percentage. A score of 92% feels like 92% of the work done — but 92% across what? The dashboards rarely segment by search mode, and when they do, the segmentation is buried in advanced reporting most clients never open. The result is a generation of practitioners who genuinely believe their work is mode-agnostic.

What Google Actually Weighs in 2024

Google’s local ranking system, as inferred from patent filings, the Google Business Profile help documentation, and observable behaviour across hundreds of audits, leans heavily on three signals: the Google Business Profile itself (categories, attributes, reviews, photos, posts), proximity to the searcher, and a relevance layer derived from on-page content and link equity. Third-party citations contribute, but they contribute as a corroboration signal — Google treats them as evidence that the entity is real and consistent, not as a primary ranking input.

The practical consequence: once a business has ten to fifteen high-quality, NAP-consistent citations on tier-one sources (Yelp, Facebook, Apple Maps, Bing Places, BBB, industry-specific directories, and the dominant data aggregators — Data Axle, Foursquare, Localeze), additional citations show diminishing returns in Google’s local pack. Audit data from 38 mid-market service businesses tracked between 2022 and 2024 shows correlation between citation count and Google local pack position dropping below r = 0.18 once businesses exceed 25 listings. That is not zero, but it is well below the threshold at which additional spend can be justified by ranking improvement alone.

How Bing Diverges From Google

Bing’s Heavier Reliance on Yelp

Bing’s local results draw disproportionately from Yelp’s data graph. This is not speculation — Microsoft and Yelp have publicly acknowledged data partnerships dating back to 2014, and the visual integration of Yelp reviews directly into Bing’s local results is observable in any test query. In categories where Yelp has dense coverage (restaurants, hospitality, personal services, automotive), a strong Yelp profile is more predictive of Bing visibility than a strong Google Business Profile is.

This has direct implications for any business whose customers skew toward Bing. The conventional dismissal — “Bing is only 6% of search, ignore it” — collapses in specific contexts. Bing is the default search engine on Windows machines, on Edge, and on a substantial portion of enterprise environments. Government contracts, B2B services targeting corporate buyers, and any vertical with an older demographic skew often see Bing market share above 15%. For those clients, neglecting Yelp is a measurable error.

Yahoo Local Listings Behavior

Yahoo Local, which retains a non-trivial residual user base in certain age cohorts, pulls from a hybrid feed that includes Yelp, Yext, and its own legacy listings database. The behaviour is inconsistent — Yahoo sometimes displays current data, sometimes data three to five years stale. The practical implication is that Yahoo Local is rarely worth direct optimisation, but it can occasionally surface outdated NAP data that contaminates the broader graph. Quarterly checks (not quarterly optimisation) are sufficient.

DuckDuckGo’s Apple Maps Dependency

DuckDuckGo, which has grown to roughly 2-3% of US search and substantially more in privacy-conscious European markets, sources its local results primarily from Apple Maps. This is an important detail: a business with a perfect Google Business Profile, comprehensive Yelp coverage, and a Bing Places listing can still be invisible in DuckDuckGo if its Apple Maps Connect entry is missing or inaccurate. Apple Maps, in turn, relies on a constellation of data partners (Yelp, TripAdvisor, Booking.com, Foursquare) and on direct submissions through Apple Business Connect.

The asymmetry here is severe. Apple Business Connect was opened to all US businesses only in 2023, and adoption among mid-market firms remains under 50% by my own estimate based on sampled audits. Businesses that complete the Apple Business Connect setup typically see DuckDuckGo and Siri visibility improve within four to six weeks, with no corresponding uplift in Google. The mode-specific nature of the gain is unambiguous.

Client Case: The Plumber Who Wasted $4K

The Original Citation Strategy

A plumbing contractor in the Midlands engaged a previous agency in early 2022 to “fix his local SEO.” The agency executed a textbook campaign: 87 citations built across general and home-services directories over three months, at a cost of approximately £4,200 including the agency retainer and listing fees. Reports were delivered, screenshots filed, completion celebrated. The client was told to expect ranking improvements within 90 days.

Why It Only Moved Bing Rankings

Six months later, the client engaged us for a second opinion because his Google local pack rankings had not moved. The audit found two problems. First, his Google Business Profile had three unaddressed category misalignments and zero photos uploaded in the prior year — issues the citation campaign had not touched because it was not designed to. Second, the citation work had concentrated on directories that fed the Bing/Yelp graph, which is why his Bing visibility had genuinely improved (from page two to position three for his primary commercial query) while Google had remained static. The client’s customer mix was 78% Google-driven, based on Google Analytics referrer data. The campaign had succeeded in moving the wrong dial.

The remediation took six weeks and cost under £900: GBP cleanup, photo upload schedule, review acquisition workflow, and three targeted citation corrections on aggregator sources. Within ten weeks the business held positions one to three for its core service queries in Google’s local pack. The original £4,200 spend was not entirely wasted — Bing rankings did improve — but the return on that investment, measured against the client’s actual customer mix, was approximately £0.22 per pound spent.

Practical Implication for Your Audit

Before any citation work begins, the first audit step should be a search-mode attribution analysis: what proportion of the client’s actual customers come from each engine? Google Analytics, server logs, call-tracking data with referrer capture, and customer surveys all contribute. Only then can a citation strategy be calibrated to the right graph. The figures presented in Table 1 confirm just how unevenly citation sources feed each search mode, and why a one-size strategy systematically underperforms in any single channel.

Table 1: Directory Citation Source Influence by Search Mode (audit data, 2022–2024, n = 47 mid-market service businesses)

The asymmetry is the point. A campaign built around the bottom half of that table will produce listings without producing rankings — a phenomenon that Harvard Business Review’s case archive repeatedly illustrates in turnaround scenarios where activity gets confused with progress.

Myth Two: More Citations Always Beats Fewer

Where the Volume Obsession Came From

Legacy Advice From 2015 Whiteboard Fridays

Volume-based citation thinking is a fossil from a specific period in local search history (roughly 2011–2016) when Google’s local algorithm did weight citation quantity more heavily and when the directory ecosystem was less saturated with low-quality listings. Industry voices on platforms like Moz, Search Engine Land, and the original Local Search Forum advocated building 50–100 citations as a baseline. That advice was reasonable then. It is not reasonable now, and the evidence has been clear since at least the 2018 algorithm updates that introduced stronger entity-recognition models.

What persists is the inertia. Agency proposals still lead with citation volume because volume is easy to sell, easy to report, and easy to compare across vendors. A proposal offering “150 citations across our premium network” looks more substantial than one offering “12 carefully selected and verified placements.” The first is worse. The second is what the data supports.

Why Aggregator Services Pushed Volume

The aggregator business model — Yext being the most prominent example, though the pattern applies broadly — depends on per-listing or per-network pricing. The economic incentive runs directly counter to the practitioner’s interest in efficient citation strategy. Harvard Business Review’s editorial criteria explicitly warn against findings that are “easily replicable by simply asking a large language model” — and yet that is precisely what aggregator-driven citation reports often are: lists generated by templates, populated by automation, and signed off without anyone asking whether each individual placement contributes anything to the client’s actual search visibility.

The Quality Threshold Data Shows

NAP Consistency Across 50 Sources

The single most important quality factor in a citation profile is NAP (name, address, phone) consistency across the sources that feed each search engine’s data graph. Audit data across 73 businesses shows that profiles with 95%+ NAP consistency on tier-one sources outperform profiles with 100+ citations and only 70% consistency by an average of 2.4 positions in the local pack. The evidence indicates that consistency dominates count.

The mechanism is straightforward: when search engines encounter conflicting NAP data across sources, the entity-resolution algorithm becomes less confident that all the listings refer to the same business. Lower confidence translates into lower trust, which translates into lower rankings. Adding more citations with inconsistent data actively harms performance because each new inconsistency increases ambiguity.

Diminishing Returns Past Tier-One Sites

Plotting citation count against ranking position across the 73-business dataset produces a curve that flattens sharply after the 15th high-quality citation and is essentially flat past the 25th. The 50th citation contributes statistically nothing. The 100th is, in many cases, a negative — because low-quality directories are more likely to scrape outdated data, propagate inconsistencies, and dilute the quality signal.

Voice Search and AI Mode Differences

Voice search and the emerging AI answer engines (Google’s AI Overviews, ChatGPT search, Perplexity, Claude’s web tooling) introduce a further wrinkle that the volume orthodoxy cannot accommodate. These modes typically retrieve a single answer or a very small set of candidates, and the retrieval logic appears to prioritise entities with strong, consistent first-party signals (the GBP, the website’s structured data, prominent first-party reviews) over breadth of citation footprint. A business with 200 citations and a thin GBP loses to a business with 18 citations and a richly populated GBP every time in voice queries.

The reasoning, as best as can be inferred from observable behaviour and from the patent literature, is that AI retrieval systems penalise ambiguity even more aggressively than traditional ranking. When the system must select one answer, it cannot afford uncertainty about the entity’s identity, and citation breadth without consistency is an uncertainty signal, not a confidence signal. As discussed in this blog post on entity resolution, the shift from ten-blue-link retrieval to single-answer retrieval changes the optimisation problem in its essence — the goal is no longer to be among the top results, but to be the unambiguous answer.

Client Case: The Dentist With 200 Listings

The Cleanup Audit We Ran

A dental practice in Manchester had spent four years and approximately £14,000 building citations through three successive agencies. The audit identified 217 live citations across 184 unique sources. Of those, 31 carried the original (now incorrect) practice phone number from before a 2020 line change, 18 listed an old address from a 2019 relocation, and 47 had the practice name slightly malformed (e.g., “Dr Smith Dental” instead of “Smith Family Dental Practice”). The practice was visible on Google’s local pack only for the most generic, low-volume queries. For competitive commercial terms (“emergency dentist Manchester,” “Invisalign Manchester”), they ranked outside the top 20.

The cleanup process took eleven weeks. Of the 217 citations, 89 were corrected, 14 were claimed and brought into NAP alignment, and 68 were either suppressed, deleted, or flagged for removal at the directory level. The remaining 46 were judged to be either inert (zero traffic, no graph influence) or actively harmful and were left alone after suppression attempts failed.

Ranking Lift After Pruning

Within fourteen weeks of cleanup completion, the practice’s average local pack position for tracked commercial queries improved from 22.3 to 6.1. Within twenty-six weeks, the practice held a top-three position for 11 of 14 tracked terms. The cleanup cost £2,400 — less than 18% of what had been spent building the original mess. A breakdown is provided in Table 2.

Table 2: Citation Cleanup Outcomes by Action Type (Manchester dental practice, 26-week tracking)

Note that the GBP cleanup, which was a single asset addressed in approximately four hours of work, contributed more measurable ranking lift than any of the citation work — a finding consistent with what every recent local search study has reported, and a finding that a related discussion in the context of profile completeness as a primary ranking signal.

What Actually Matters for Citation Strategy

The Tier-One Directories Worth Your Time

Mode-Specific Priority Mapping

Stripped of the noise, citation strategy reduces to a small number of decisions made carefully rather than a large number made carelessly. The tier-one set varies slightly by vertical and geography, but for a UK-focused service business in 2024 the priority order generally runs: Google Business Profile, Bing Places, Apple Business Connect, Yelp, Facebook Business, the dominant data aggregators (Foursquare, Data Axle), the BBB or its UK equivalent (Trustpilot, depending on vertical), and two to four industry-specific directories. That is roughly twelve to fifteen placements, each verified, each NAP-aligned, each updated when business data changes.

According to a study available here on placement quality versus quantity, the marginal value of the 16th citation falls below the cost of placing it for the great majority of mid-market businesses. The exceptions — multi-location franchises, businesses operating in densely competitive verticals, and entities with active reputation-management requirements — justify additional placements but do so for reasons other than ranking lift.

Industry-Vertical Adjustments

Vertical-specific directories produce disproportionate value when they are genuinely authoritative within the industry’s information ecosystem. Healthgrades and Vitals carry weight for medical practices in a way that no general directory replicates. Avvo and Justia matter for legal. TripAdvisor and Booking.com matter for hospitality in ways that affect not only direct bookings but also the Apple Maps and DuckDuckGo graphs through Apple’s data partnerships. The principle: a vertical directory earns priority placement only if it is referenced by users in the vertical or fed into the search engines’ data graphs. Most niche directories meet neither criterion.

Cross-referencing Table 3 reveals how dramatically the priority list shifts across three representative verticals, and why generic citation packages systematically misallocate effort.

Table 3: Top Three Citation Priorities Beyond GBP/Bing/Apple by Vertical

A Repeatable Quarterly Citation Workflow

The work that genuinely moves rankings, sustained over time, fits into a quarterly rhythm rather than a campaign-based push. The cadence reflects the rate at which directory data drifts, business information changes, and the search engines refresh their graphs.

The first month of each quarter is audit. A scan across the tier-one set (Whitespark, BrightLocal, or Moz Local all serve adequately for the discovery phase, though the data each returns differs enough that cross-checking is sensible) identifies new inconsistencies, scraped duplicates, and newly emerged listings. The audit also reviews search-mode attribution: has the client’s customer mix shifted? A B2B firm that began winning more public-sector contracts may have moved from a 90/10 Google/Bing split to 70/30, and the citation priorities should follow.

The second month is correction. The temptation here is to delegate to automation, but evidence indicates that automated cleanup tools have a non-trivial error rate (estimates range from 8% to 14% based on published comparisons, though the figure varies by tool and by source-directory cooperation). For tier-one citations, manual verification remains worth the time. For tier-two and below, automation’s error rate is acceptable because the ranking impact of those citations is already low.

The third month is enhancement. This is where the work shifts from defensive (keeping data consistent) to offensive (adding features that strengthen the entity signal). New photos to the GBP, new posts, new attributes as Google releases them, new review-acquisition outreach, new structured data on the website. The enhancement phase, more than any other component of the workflow, is what produces continuing ranking improvement after the initial cleanup baseline is established.

One reflective remark from years of running this cycle: the clients who see the largest sustained gains are not the ones who invest the most in citations, but the ones who treat the GBP as a living asset rather than a set-and-forget profile. The citation work is the foundation; the GBP is the building. Pouring more concrete after the foundation is set does not raise the building any higher.

The methodological caution from Pew Research Center bears repeating in this context — aggregate metrics regularly conceal subgroup divergence. A “92% citation health score” tells a practitioner almost nothing useful. The score that matters is consistency on the specific sources that feed the specific search modes the specific customer base actually uses. That score is harder to calculate, harder to sell, and harder to celebrate in a quarterly report. It is also the only one that correlates with the ranking outcomes clients pay for.

Two practical implications follow directly from the analysis above and deserve to be stated as decisions rather than as observations. First, any agency engagement or in-house programme that begins with a citation count target — “we will build 75 citations” — is starting from the wrong premise and should be restructured before any spend is committed; the correct starting question is which search modes drive the client’s revenue and which sources feed those modes, and the citation count falls out of that analysis as a consequence rather than as a goal. Second, for any business that has already been through one or more legacy citation campaigns, the highest-return next step is almost certainly a cleanup audit rather than additional building; the Manchester dental case is not unusual, and the pattern of legacy-debt-as-ranking-anchor appears in roughly six out of ten audits conducted on businesses with more than three years of citation history. Third, and more strategically, the rise of AI-driven search modes is shifting citation work from a breadth game to a precision game faster than agency pricing models have adjusted, which means that the next eighteen to twenty-four months will likely produce a meaningful gap between practitioners who restructure their service offerings around mode-specific entity strength and those who continue selling volume — a gap that will be visible in client retention rates well before it shows up in industry commentary.