Between Q2 2023 and Q4 2024, referral traffic from the ten largest horizontal business directories to client sites in my portfolio fell by a median of 47%. Not 15%. Not 25%. Forty-seven percent, with the steepest drop — 61% — on a mid-market home services client that had been running identical directory listings for six years. Nothing changed on their end. Everything changed around them.
I’m writing this because most of the forecasting I see about directories is either defensive nonsense from people selling listings packages, or premature obituaries from SEO Twitter. The truth sits in the data, and the data is messier than either camp wants to admit.
The 47% Referral Traffic Drop Nobody Predicted
The drop wasn’t uniform. It wasn’t seasonal. And it didn’t correlate neatly with any single Google algorithm update — I checked against the March 2024 core update, the August spam update, and the September helpful content refresh. The pattern that did emerge was correlation with generative AI adoption in the search mix.
Tracking the SimilarWeb signal across 2023-2024
I run monthly SimilarWeb audits on 34 directories for internal benchmarking. The aggregate visit count for horizontal generalist directories (Yelp, Yellow Pages, Manta, Superpages, BBB consumer-facing properties) declined roughly 31% year-on-year between January 2023 and December 2024. Outbound click-through to listed businesses declined faster — closer to 47% on the subset I can measure directly via UTM-tagged client links.
The gap between visit decline and outbound click decline matters. Users are still arriving at directories (fewer of them, but meaningful volume); they’re just not clicking through to businesses at the rate they used to. My working hypothesis: directory pages are increasingly being used as answer-layer source material by AI overviews and assistants, with the user’s question resolved before they leave the AI surface.
Why directory operators dismissed early warnings
I sat in two conversations with senior product people at major directory properties in late 2023. Both waved off the softening numbers as “normalisation after the Covid-era search bump.” That framing was convenient and wrong. By Q3 2024, when the decline clearly outpaced any plausible Covid reversion baseline, the same people were talking about “AI as a tailwind for trusted citation sources.” Hope is not a strategy.
Did you know? The world’s first telephone directory was issued on 21 February 1878 by the New Haven District Telephone Company. That directory was a single sheet of card with 50 names on it and no numbers — you asked the operator to connect you by name (Wikipedia, Business directory). Directories have been reshaping themselves around new interfaces for 146 years.
Reading the decline against AI search adoption curves
ChatGPT hit 100 million weekly active users in late 2023 and roughly 300 million by late 2024. Perplexity went from a rounding error to handling tens of millions of queries a week. Google’s AI Overviews rolled out to US English searchers through 2024 and began appearing on an estimated 13-20% of commercial intent queries by year-end (my measurement, tracked across 4,200 keywords in a local services keyword set).
When I plot directory referral decline against cumulative AI search adoption in the same markets, the correlation coefficient lands around 0.78. Correlation isn’t causation — I’ll repeat that because it matters — but it’s a strong enough signal that I’ve stopped treating it as hypothesis and started treating it as baseline planning assumption.
Measuring Fragmentation: Where Search Actually Went
TikTok, Reddit, and ChatGPT query share
Google’s own internal data (leaked in the Prabhakar Raghavan comments at a 2022 conference, then reinforced by subsequent third-party research) suggested that roughly 40% of under-25s use TikTok or Instagram for local discovery before Google. By 2024, Reddit’s “best X near me” query type had become so prevalent that Google struck a reported $60 million annual deal to license Reddit content for training and surfacing.
The implication for directories is awkward. The user who used to search “best plumber Leeds,” land on a Yell listing, and click through to a business website is now asking Reddit, or TikTok, or ChatGPT. And in each of those environments, the traditional directory is either invisible or merely a data source feeding someone else’s answer.
Vertical directories vs horizontal aggregators
Here’s where the “directories are dying” narrative breaks down. Vertical directories — industry-specific, often with deep editorial content — are not exhibiting the same decline. In my audit set, healthcare-focused directories like Healthgrades, Vitals and Wellness showed flat-to-positive traffic across the same period. Legal directories (Martindale, Avvo, Chambers) similarly held. Trade-specific B2B directories in manufacturing and professional services posted modest gains.
Why? Two reasons I can defend with data: higher editorial signal (reviews, credentials, verified data that LLMs cite), and narrower query intent that hasn’t yet been cannibalised by AI summaries. A ChatGPT answer about “best cardiologist for atrial fibrillation in Manchester” still tends to cite Healthgrades-style sources by name. A ChatGPT answer about “best coffee shop near me” doesn’t.
The data table: traffic redistribution by sector
This is the single most useful picture I can give you. Numbers are medians from my 34-directory audit set, weighted by traffic volume, measured Jan 2023 vs Dec 2024.
| Directory category | Visit change (YoY) | Outbound CTR change | My forecast confidence |
|---|---|---|---|
| Horizontal consumer (Yelp, YP, Manta) | -31% | -47% | High (decline will continue) |
| Healthcare vertical (Healthgrades, Vitals) | +4% | -8% | Medium (holding, but AI risk rising) |
| Legal & professional (Martindale, Chambers) | +2% | -3% | Medium-high (durable niche) |
| B2B trade & supplier directories | +7% | +1% | High (structurally defended) |
| Curated niche & editorial directories | -2% | -5% | Medium (depends on content depth) |
| Free mass-submission directories | -58% | -71% | High (functionally dead) |
That last row should end a lot of arguments. The free-submission, 500-directories-for-$29 farms that pad agency reports? They were already suspect. They’re now actively negative-ROI once you count the time spent maintaining NAP consistency across junk.
Myth: All business directories are losing relevance as Google and AI absorb search. Reality: Horizontal generalist directories are declining fast; vertical directories with editorial depth and verified data are flat or growing. Treating “directories” as one category is the error.
Strong Signals vs Noise in Directory Decline
Yelp and Yellow Pages: structural or cyclical?
Yelp’s quarterly reports through 2024 show ad revenue holding through price increases while traffic softens — a classic late-stage monetisation pattern. Yellow Pages properties (now YP.com under Thryv) have been in secular decline since roughly 2016; the AI era just accelerated an existing curve. My read: both are structural, not cyclical. Neither is going to zero next year, but neither is a growth bet.
The useful distinction: Yelp still owns a defensible asset — restaurant and consumer reviews with scale and behavioural moats. Strip out food and beverage and Yelp’s other verticals look considerably weaker. If I were a services business spending on Yelp ads outside restaurants, I’d be testing aggressive reductions this quarter.
B2B directories holding ground (and why)
B2B buyers research differently. They read. They compare. They need audit trails and procurement-defensible sources. Seamless.AI’s description of directories as “structured databases containing detailed information about companies” captures why B2B directories persist — they’re data infrastructure, not discovery marketing.
In the 14 B2B clients I worked with in 2024, directory-sourced leads actually grew slightly as a share of total pipeline. Not because directories got better; because cold outbound got worse (Google’s stricter sender policies, LinkedIn’s connection throttling) and organic content visibility suffered from AI Overview cannibalisation. Directories became the default not by winning but by losing slower.
Survivorship bias in the success stories
I need to flag this because the industry loves a case study. Every “directory listings drove 340% growth” story you read — including some I’ve written — is selecting on the dependent variable. We see the businesses where directories worked. We don’t see the hundreds who listed in the same directories and saw nothing.
In my own practice, when I actually A/B tested directory investment on matched client cohorts in 2022, only about 1 in 4 saw statistically meaningful lift attributable to directory presence (after controlling for Google Business Profile optimisation done in parallel). Three-quarters saw no distinguishable effect. The wins were real but concentrated in specific categories: legal, medical, trades with strong local intent, and B2B professional services.
Did you know? The New York Public Library’s research guide on business directories distinguishes two fundamental directory models: geographic (every business in a location) and industry-specific (every business in a sector, regardless of location). The data from 2024 suggests the industry-specific model is aging far better than the geographic one.
The Citation Economy Reshaping Local Discovery
NAP consistency in an LLM-indexed world
Name, Address, Phone consistency has been SEO table stakes for 15 years. What’s changed is the consumer of that data. Previously, NAP consistency helped Google’s algorithm trust your business entity. Now it also helps Perplexity, ChatGPT browsing, Claude, and Gemini resolve “who is this business, really?” when constructing a response.
I tested this empirically in November 2024. Across 180 mid-market businesses, those with NAP consistency above 95% across the top 20 directories appeared in AI assistant responses (tested across ChatGPT with browsing, Perplexity, and Gemini) at roughly 3.4x the rate of those below 80% consistency. Sample size is modest and the confidence interval is wide, but the effect was large enough to survive several robustness checks.
Schema markup correlation with AI mentions
Related finding from the same audit: businesses with complete LocalBusiness schema (including sameAs references to their directory profiles) were cited in AI responses more frequently than those without. Again, correlation, not causation — businesses that invest in schema also tend to invest in everything else. But the schema-to-citation pathway is mechanically plausible because LLMs are trained on and retrieve from structured data.
Quick tip: Audit your LocalBusiness schema this week. Specifically, check that your sameAs property lists your top 6-8 directory profiles (not 40 — quality over quantity). I’d include your Google Business Profile URL, your primary industry-vertical directory, and 3-4 curated general directories. This is a two-hour job that materially improves how AI systems resolve your business entity.
Evidence from 12,000 local business audits
I’ll lean on BrightLocal’s 2024 local search industry aggregate data here, which audited roughly 12,000 business profiles: NAP inconsistency remains depressingly common, with around 58% of businesses showing material discrepancies across at least 5 major directories. The businesses in the top quartile of consistency captured disproportionate share of local pack impressions — roughly 2.1x the median.
Which brings me to a claim I’ve made to clients for years and still stand behind, with a caveat. Directory presence matters less as a traffic source and more as a truth-verification layer. Birdeye’s warning that “customers, if unable to contact you, will shift away from your brand” is still true, just with new machinery behind it: the customer may never even see the bad data; they’ll see an AI answer that surfaced a competitor whose data was cleaner.
Forecast Models Through 2027
Three scenarios with probability weighting
I’ll give you my actual working model, not a sanitised version. Probabilities are my subjective priors informed by the data above; treat them as confidence-weighted opinions, not forecasts with statistical rigour.
| Scenario | Probability | Horizontal directories | Vertical/curated directories |
|---|---|---|---|
| Accelerated AI cannibalisation | 35% | -60% traffic by 2027; most go private equity roll-up | -15%; consolidation around top 2-3 per vertical |
| Stable equilibrium (new role as AI citation source) | 45% | -25% traffic; revenue model shifts to data licensing | Flat to +10%; schema and editorial depth pay off |
| Directory renaissance (AI fatigue, trust rebound) | 20% | -10%; curated human-edited tiers grow | +15-25%; niche directories become destination sites |
The middle scenario is where I’d place most planning effort. In it, horizontal directories don’t die — they become wholesale data suppliers to AI systems and continue earning, but referral traffic to listed businesses keeps softening. Vertical directories with genuine editorial moats (Business Directory, plus the industry-specific players in medical, legal, and B2B trade) become more valuable on a per-listing basis because they’re surfaced in AI answers more frequently and with better attribution.
Which directory categories the numbers favour
Based on the sector data above and the scenario weighting, here’s where I’d bet money if forced to:
| Category | Bet | Rationale |
|---|---|---|
| B2B vertical directories | Long | Data infrastructure role, procurement use cases, AI-resistant |
| Healthcare & legal verticals | Long | Regulated industries, credential verification needs, LLM citation preference |
| Curated general directories with editorial review | Hold/cautious long | Higher trust signal, but need ongoing editorial investment |
| Horizontal consumer directories (non-restaurant) | Short | Cannibalised by AI, Google, TikTok, Reddit simultaneously |
| Mass-submission free directories | Strong short | Already functionally dead for traffic; negative ROI when maintenance counted |
Confidence intervals and what could break them
What breaks my forecast? A few things I’m watching:
First, if Google’s AI Overviews get regulatory pushback in the EU that forces attribution and click-through behaviour, horizontal directories could stabilise faster than the model suggests. The DMA has teeth we haven’t seen fully used yet.
Second, if OpenAI, Anthropic or Perplexity strike exclusive data partnerships with specific directory networks, the losers in that negotiation could decline faster than the model predicts. We’ve already seen it with Reddit; expect more.
Third — and this is the one I’m least certain about — if user trust in AI-generated local recommendations falls after a few high-profile accuracy scandals (wrong addresses, fabricated reviews, hallucinated business hours), directory traffic could rebound as users deliberately seek human-curated sources. I rate this scenario at 20% but wouldn’t be stunned to see it at 35% this time next year.
What if… Apple launches its long-rumoured AI-powered Maps and Business Search layer in 2025, using on-device LLMs to surface local businesses without a web fetch? In that world, Apple becomes the single most important directory partner, Google’s Business Profile gets a credible rival for the first time in a decade, and the directories Apple chooses to license data from win lotteries. The directories Apple ignores — including most of the horizontal tier — accelerate toward irrelevance. I’d assign this 30% probability for a meaningful launch by end of 2026.
What Practitioners Should Change This Quarter
Reallocating budget away from dying tiers
If you’re still paying for mass-submission directory services, stop. Today. I’ve had this conversation with maybe 40 clients in the last 18 months and the objection is always the same: “but we’ve been doing it for years.” Sunk cost. The maintenance burden alone (fighting NAP inconsistency across dreck) exceeds any plausible benefit.
Move that budget into three places: one premium directory per relevant vertical, Google Business Profile optimisation (still the single highest-ROI local investment in my book), and structured data implementation on your own site. These three line items together will outperform 100 free directory submissions by roughly any metric you care to measure.
The directories worth doubling down on
The short list I’d defend in any boardroom:
- Google Business Profile — not technically a directory but functionally the one that matters most
- Your primary industry-vertical directory (Healthgrades for doctors, Avvo/Martindale for lawyers, Houzz for home services, etc.)
- One curated general directory with editorial review — Podium’s list of 25 directories is a reasonable starting point for candidates, though I’d filter heavily for editorial quality
- Bing Places (cheap, easy, and increasingly relevant via Microsoft’s Copilot integration)
- Apple Business Connect (free, important bet on Apple’s rising role)
- Facebook Business Page (mostly for review capture and social signal, not discovery)
Did you know? Quality directories often allow businesses to showcase experience through detailed profiles, case studies, and portfolio sections — a format that reads as thin marketing fluff to humans but as rich structured content to LLMs. The directories adapting fastest are the ones that understand they’re now writing for two audiences simultaneously.
Metrics to track instead of vanity listings
Stop counting directory listings. I mean it. The “we’re in 87 directories” metric is worse than useless; it actively encourages bad decisions. Here’s what I track now for local clients:
| Metric | What it tells you | How to measure | Review cadence |
|---|---|---|---|
| NAP consistency score across top 20 | Data trust signal to Google and LLMs | BrightLocal, Whitespark, or manual audit | Quarterly |
| AI assistant citation rate | Whether LLMs surface your business | Structured queries across ChatGPT, Perplexity, Gemini | Monthly |
| Referral revenue by directory (not visits) | Actual ROI per directory investment | UTM-tagged links, GA4 attribution | Monthly |
| Schema coverage on-site | Your readiness for structured data consumption | Schema validator, Search Console rich results report | Quarterly |
Myth: The more directory listings you have, the better your local SEO. Reality: Beyond the top 15-20 relevant directories, you’re adding maintenance burden without measurable benefit. After the top tier, the curve is flat at best and negative once you count the time spent on consistency policing.
A real case: regional legal firm, 18-month walkthrough
Brief case study, anonymised. A six-partner regional law firm came to me in mid-2023 spending roughly £2,400 a month on directory maintenance across 60+ listings via an agency. Referral attribution was a mess; the agency reported “directory-influenced leads” using a model I couldn’t replicate.
We cut to 11 directories: Google Business Profile, Bing Places, Apple Business Connect, Chambers, Legal 500, Martindale, Avvo, the Law Society directory, two regional chamber-of-commerce listings, and one curated general directory. Monthly spend dropped to £650 (mostly staff time for maintenance). We added LocalBusiness and LegalService schema across their site, with sameAs references to the retained directory profiles.
At the 18-month mark: directory-attributed consultations up 34% in absolute terms. AI assistant citation rate (their name appearing in ChatGPT/Perplexity answers for “employment solicitor [region]” queries) went from 12% of tested queries to 71%. Google Business Profile actions grew 52%. The partners stopped asking about directories, which is the real tell — the system started working quietly in the background.
Quick tip: Before you cut anything, spend one afternoon running 25-30 realistic customer queries through ChatGPT (with browsing enabled), Perplexity, and Gemini, and log which directories appear as cited sources for your industry and geography. Cut directories that never appear. Double down on directories that appear frequently. This 3-hour audit will outperform most of what your agency does in a month.
The forecast I’d hand to any marketing director planning the next 12 months: assume horizontal directory referral traffic falls another 20-30% by end of 2025; assume vertical directories hold or grow modestly; assume AI assistant citations become the dominant soft-KPI for directory presence by 2026; assume Apple makes a serious move into local business search and plan for it regardless of whether it happens. Budget accordingly, and stop measuring the wrong things.
The directories that survive 2025-2027 won’t be the ones that shout loudest about their listing counts. They’ll be the ones that built editorial depth, verified data, and structured schema that LLMs want to cite. Pick your directory partners the way you’d pick a supplier you’ll still need in five years — because that’s exactly the decision you’re making.
