This is a case study report on an experiment I ran in early 2024. Two niche edit orders placed with FATJOE in January and February 2024, totalling £2,150.40 across six placements, plus PR Fire researched and rejected as a category companion. The placements were tracked for fifteen months. The report explains what was bought, what got delivered, what the metrics showed, why the mechanism predicts the outcome, and what the same budget would have produced reallocated to Editorial Selection-building activity.
The piece is grounded in invoiced evidence with screenshots available on request. Vendor companies (FATJOE, PR Fire) are named factually as the parties to the transactions. The motoring law client whose URLs appear on five of the six placements is anonymised here as a UK motoring law client at the client’s preference; the seostrategy.co.uk placement is named explicitly because it is my own site. No vendor is characterised; the analysis is mechanistic and applies to the categories of paid-placement service these vendors sell rather than to the vendors themselves.
The reason to publish this report is the same reason I cited Lily Ray’s 220-site Mount AI failure dataset in Footprint vs Fingerprint and Aaron Haynes’s 0-of-300 press release citation finding in Editorial Selection: the industry’s published evidence base on paid-placement outcomes is heavily weighted toward vendor case studies celebrating short-term wins. Practitioner-side reports of what these tactics actually produce across fifteen-month horizons are rare. The asymmetry produces a buying decision that looks more rational than it is. This is one practitioner’s contribution to fixing the asymmetry.
The setup — why I ran the experiment
I have been doing SEO since 2005 and consulting full-time since 2010. By January 2024 I had been observing AI retrieval shifts in my client work for about eighteen months and had started forming the thesis that would later become the Editorial Selection framework: the mechanism that produced an inclusion was now mattering more than the inclusion itself.
Client work creates pressure to test that thesis empirically rather than only argue it from first principles. Two of the clients I was working with in January 2024 were under commercial pressure to produce visible link-acquisition progress quickly. One was a UK motoring law firm operating in a competitive search environment; the other was my own consultancy site, seostrategy.co.uk, which I had been treating as a deliberately constrained test bed for off-page tactics I was evaluating before recommending or rejecting them for clients.
The choice was between two postures. Argue the thesis from theory and decline to run the test, recommending only Selection-mechanism activity (journalist outreach, proprietary data, conference participation) on the client side. Or run the test under controlled conditions, document the outcomes rigorously, and produce the evidence that would either confirm or refute the thesis. I chose the second posture because I think the first posture is intellectually weaker when the empirical question is answerable.
The experiment ran in two cohorts. Cohort one: FATJOE niche edit order FJ0124591698 on 30 January 2024, £806.40, two placements. Cohort two: FATJOE niche edit order FJ0224061044 on 18 February 2024, £1,344.00, four placements. Six placements total, £2,150.40 total spend, all delivered within SLA by mid-March 2024. PR Fire was researched in parallel during this window as a press release distribution alternative; the research outcome was a decision against purchase based on emerging evidence (Aaron Haynes’s analysis was published in preliminary form during 2024) that the press release category would produce close to zero AI citation value regardless of distribution scale. The PR Fire research is included in this report as the category companion case — the same diagnostic applied to the adjacent paid-placement category without consuming the budget to confirm it.
What I bought from FATJOE — the six placements
The two FATJOE orders produced six placements across five host sites. Anchor text was specified at the order stage; target URL was specified at the order stage; host site selection was managed by FATJOE from their marketplace inventory. All six placements went live within the standard FATJOE timeline. The deliverables matched the order specifications.
Cohort one (30 January 2024 order, FJ0124591698, £806.40):
Placement 1: anchor Drink Driving Solicitors pointing to the client’s drink-driving offence page, placed on lifeinabreakdown.com in an article titled 6 Things to Think about When Buying a Car, listed at DR 60. The host article is a consumer car-buying piece; the inserted link is a commercial motoring solicitor anchor. Topic adjacency: weak. The article does not naturally discuss drink driving offences.
Placement 2: anchor Dangerous Driving Charge pointing to the client’s dangerous driving offence page, placed on brightonjournal.co.uk in an article titled Cruising Along the Coast: Planning a Road Trip and Staycation from Essex to Brighton with Family, listed at DR 59. The host article is a family staycation piece; the inserted link is a commercial motoring solicitor anchor. Topic adjacency: very weak. A family staycation article carrying a Dangerous Driving Charge link reads, to a human or to a sufficiently capable retrieval system, as paid insertion.
Cohort two (18 February 2024 order, FJ0224061044, £1,344.00):
Placement 3: anchor Best Drink Driving Solicitors London pointing to the client’s London legal services page, placed on menswearstyle.co.uk in a 2021-dated article titled learning-to-drive/9545, listed at DR 53. Host article: men’s fashion site, learning-to-drive content. Topic adjacency: very weak. The slug format suggests programmatic article generation.
Placement 4: anchor SEO Agency Southampton pointing to seostrategy.co.uk’s Southampton consultant services page, placed on bbntimes.com in an article titled Top Digital Marketing Mistakes Businesses Need to Avoid, listed at DR 72. Topic adjacency: defensible — an SEO agency link inside an article on digital marketing mistakes is at least thematically related. This is the only one of the six placements where the topic adjacency does not signal paid insertion on inspection. It is also the only placement on my own site rather than the client’s. The contextual fit was the reason I selected this placement specifically for the seostrategy.co.uk test.
Placement 5: anchor Motoring Offence Solicitors pointing to the client’s homepage, placed on salonprivemag.com in an article with the slug iabout-keeping-a-car-for-a-long-time, listed at DR 58. The slug’s stray i prefix and the host article’s structure both suggest paid-placement origination of the host content itself. Topic adjacency: weak. A luxury lifestyle magazine hosting a motoring offence solicitor link is not editorially natural.
Placement 6: anchor Drug Driving Solicitors pointing to the client’s drug-driving offence page, placed on intheplayroom.co.uk in an article titled Helpful Pieces Of Advice When Buying A Family Car, listed at DR 54. Host article: parenting site, family car article. Inserted link: drug driving solicitors. Topic adjacency: very weak. This is the placement that most obviously fails any plausibility test as editorial selection.
The six placements share a structural pattern. All target commercial keywords with exact-match anchors. All sit on host sites whose article inventory shows the same paid-placement footprint pattern when audited at the host level. All carry the four properties of Placement as defined in Editorial Selection: transactional inclusion, evaluative criteria set by the buyer, partial irreversibility from the buyer’s perspective, and signal opacity engineered to look like editorial citation. None could be defended as Editorial Selection events under the four-property test.
Why I researched PR Fire and chose not to buy
PR Fire is a UK-based press release distribution service marketing itself as the UK’s number one press release distribution platform. The service offers distribution to a syndication network presenting itself as news media. Pricing in early 2024 started at approximately £95 per release with bundle options for higher volumes. The distribution network includes a mix of news aggregators, journalist databases, and news-styled host sites that accept syndicated content via the syndication agreement rather than via independent editorial commissioning.
I researched the service during the FATJOE experiment window because the underlying mechanism is similar in structure (paid placement onto host sites that present as editorial) but operates in an adjacent category (press release distribution rather than niche edits). The diagnostic question was whether the syndication-network category produced different outcomes from the niche-edit-marketplace category despite the surface differences.
The research did not proceed to purchase because by early 2024 the empirical evidence on press release citation in AI systems was already emerging. Aaron Haynes’s analysis, eventually published in its full form in 2026 with the finding of zero out of three hundred press release citations across three hundred platform-query combinations, was circulating in preliminary form across the SEO research community during 2024. The directional evidence was strong enough that committing budget to confirm the conclusion would have produced documented failure rather than incremental information. The decision was to spend the equivalent budget elsewhere.
The PR Fire research is included in this report because the diagnostic logic applies to the category, not specifically to PR Fire. PR Newswire, Business Wire, EIN Presswire, eReleases, and adjacent services operate similar mechanisms with similar predicted outcomes in AI retrieval systems. The specific vendor whose homepage produces results on the pr fire search query in May 2026 is not the analysis. The analysis is that press release distribution to syndication networks is structurally Placement in the four-property sense, and that AI retrieval systems are trained to discount the syndication footprint specifically — the Haynes finding documents this empirically. Buyers in the category should run the five-question pre-purchase test on the category before committing to any specific vendor within it.
What got delivered vs what changed
FATJOE delivered the product they sold. All six placements went live, indexed within the expected window, and remained live throughout the fifteen-month tracking period. The links carry forward as PageRank-eligible edges in any analysis that counts links neutrally. The vendor’s service-level commitments were met. No part of this report is a complaint about service execution.
What did not happen is the visibility outcome the buying decision was implicitly investing in. Tracked across fifteen months on the client side and on seostrategy.co.uk, the six placements contributed close to zero on the metrics that determine commercial outcome. Short-term DR shifts: marginal positive in the first ninety days, consistent with the standard pattern of link metrics responding to indexed backlinks. Long-term ranking impact on the commercial keywords the anchors targeted: indistinguishable from background variance. AI citation frequency in ChatGPT, Perplexity, and Google AI Overview testing across the fifteen-month period: zero attributable to these placements specifically. Lead quality and conversion metrics on the targeted pages: no measurable difference attributable to the placement programme versus the same pages’ baseline.
The seostrategy.co.uk placement on BBN Times is the easiest to track precisely because it is on a single page on a single domain I control fully. The Southampton consultant page received the inserted link; the link is one of approximately three hundred external references to that page across the tracked period; the page’s organic traffic, AI citation rate, and conversion metrics show no perturbation correlated with the placement’s go-live date. The placement is one signal among many that no individual signal is doing measurable work.
The client side is harder to attribute cleanly because the motoring law domain has multiple visibility initiatives running concurrently. What can be said is that the five client placements did not produce measurable lift on any of the targeted commercial keywords, did not produce attributable lead inquiries to the targeted page URLs, and did not produce AI citation appearances during repeated testing across the fifteen-month period. The placements exist in the link graph; their contribution to commercial outcome is below the detection threshold of the measurement approach used.
This is the outcome the mechanism predicts. The next sections explain why.
Why the mechanism predicts this outcome
The full mechanism is documented in Editorial Selection and need not be repeated here in detail. The short version follows.
AI retrieval systems weight sources by the inclusion mechanism, not by the inclusion’s surface appearance. A placement that looks editorial but was produced by commercial transaction reads, to a system trained to distinguish, as Placement. Three published evidence points anchor the asymmetry. The University of Toronto AI Citation Study (September 2025, 13 industries): 92.1% of Google AI Overview citations come from earned editorial coverage. The Muck Rack Generative Pulse analysis (July-December 2025, over one million links): 82% of AI citations from earned editorial sources. The Aaron Haynes finding: zero out of three hundred press release citations across three hundred platform-query combinations. Three independent studies, three independent methodologies, all converging on the conclusion that paid-placement mechanism content earns near-zero AI retrieval value.
The Mount AI dataset Lily Ray published on 13 May 2026 (220+ sites, 54% with 30%+ peak traffic loss, 39% with 50%+ loss, 22% with 75%+ loss) is the strongest single body of evidence that the failure plays out across eighteen-month horizons rather than six-month horizons. Sites whose programmes were celebrated in case studies during their rapid growth phase produce the diagnostic Mount AI traffic curve in the year that follows. The placement programme documented in this report was small enough (six placements across two sites) that it did not produce a site-level Mount AI shape on either domain. The mechanism still operates: each individual placement consumes budget without contributing to the compounding asset that drives durable visibility.
The framing reviewer feedback on the Editorial Selection piece pushed for, and I now adopt explicitly, is failure to compound rather than penalty risk. The 2026 reality is that paid-placement mechanism content is mostly discounted by AI retrieval systems rather than penalised. The penalty-fear framing is the older industry conversation. The compounding-failure framing is the operationally relevant one. The reason to decline a Placement opportunity in 2026 is not that it will hurt the buyer; it is that the spend does not produce an asset that compounds, and the opportunity cost of the misallocated budget is what the same budget would have produced in Selection-building activity instead.
The five-question test, applied retrospectively
The Editorial Selection guide includes a five-question pre-purchase test designed to be run before any paid-visibility spend. The test was developed in March 2026 based on the cumulative practitioner experience that included these six placements as direct inputs. Running the test retrospectively on the six placements is a useful diagnostic exercise.
Question 1 — Mechanism. Is the inclusion produced by Editorial Selection or Placement? All six placements: Placement. They were commercial transactions where the host site’s choice to publish was determined by the FATJOE marketplace agreement, not by independent editorial judgement. Categorical fail on Question 1 across the entire portfolio.
Question 2 — Discount. What is the realistic AI-retrieval value? The Haynes 0/300, Toronto 92.1%, and Muck Rack 82% data points predict near-zero AI citation value for niche-edit-mechanism content. The fifteen-month tracking confirmed the prediction. Categorical fail on Question 2.
Question 3 — Substitution. What would the Selection-mechanism version cost? For the client placements: one journalist relationship at a motoring trade publication, pitched on the basis of the firm’s case law expertise, would have produced over twelve months approximately the same surface visibility outcome as the five paid placements, at a time cost rather than a cash cost. For the seostrategy.co.uk placement: one contributed article to a UK SEO publication, pitched on the basis of the consultancy’s specific frameworks (which by then included the early versions of CITATE and the AI Discovery Stack), would have produced higher-quality visibility than the BBN Times placement. Both substitution alternatives existed and were declined in favour of the experiment.
Question 4 — Compound. Does each placement contribute to a compounding asset? No. Each is a one-off transaction whose visibility exists at the moment the link is live and does not propagate into adjacent editorial coverage of the buyer entity. Categorical fail on Question 4.
Question 5 — Opportunity cost. What else would the £2,150.40 have produced? Reallocated to Selection-building activity, the budget would have funded one full day per month of journalist relationship-building across twelve months, or two pieces of proprietary data publication and pitching, or attendance and speaking participation at two UK SEO industry conferences. Each of those alternatives produces compounding visibility assets. The Placement alternative produced six one-off transactions with near-zero compounding value.
Five out of five categorical fails across the portfolio. The placement that scored least poorly on the surface inspection — the seostrategy.co.uk BBN Times placement with defensible topic adjacency — still failed all five questions of the structural test. The five-question test is a more reliable diagnostic than surface adjacency, and the experiment confirmed it.
What the £2,150 would do today in Selection-mode
The reallocation calculation is the practical takeaway. The same budget applied to Selection-mechanism activity in 2026 would purchase a substantially different mix.
One day per month of dedicated journalist outreach across twelve months, working a list of fifteen to twenty UK motoring law and SEO trade journalists and analysts, would cost approximately the same as the FATJOE programme at a comparable practitioner day rate. The expected output: three to five established journalist relationships of the kind that produce repeat quoting over multi-year horizons, plus four to six initial editorial mentions during the year as the relationships develop. Each of those mentions is a Selection event scoring cleanly on the four-property test; each contributes to retrieval gravity in a way the FATJOE placements do not.
Alternatively, two pieces of proprietary data publication targeting the client’s actual case law experience — for example, a quarterly analysis of regional drink-driving conviction trends in the client’s Hampshire / South coast catchment, or a longitudinal piece on procedural changes in motoring offence prosecutions — would cost roughly the same to produce as the placement programme. Each piece would be pitchable to multiple journalists, citable in trade coverage for years, and discoverable to AI retrieval systems as the kind of evidence-anchored proprietary content the Footprint vs Fingerprint framework defines as fingerprint-grade.
Or, applied to the consultancy side: one annual sponsored research project with a UK SEO industry trade body, producing a named co-authorship credit and the kind of cross-source citation that compounds across the citation network. The co-author byline is a Selection event. The research is referenced for years by other practitioners writing on adjacent topics. The relationship with the trade body produces further Selection opportunities downstream.
Each of these reallocations produces a compounding asset. The £2,150.40 placement programme produced six transactions whose contribution to the visibility outcome is below the detection threshold of standard measurement. The opportunity cost is real and is the largest practical lesson from running the experiment.
Honest close
I have been doing SEO for twenty years and I bought six niche edits in early 2024. The buy was not a lapse in understanding. I knew the thesis the buy was testing, and I knew the directional evidence predicted what the experiment subsequently produced. The reason I ran the experiment anyway was that the question is more answerable through documented practitioner testing than through theoretical argument, and the published evidence base on the buyer-side outcomes of paid-placement programmes was, and largely still is, dominated by vendor case studies rather than by practitioner reports.
This report is the practitioner-side contribution to the evidence base. £2,150.40 across six placements, fifteen months of tracked outcomes, zero attributable AI citation lift, marginal short-term DR movement that did not persist as commercial value. The mechanism predicts the outcome. The framework that names the mechanism is Editorial Selection. The companion framework that names the on-page side of the same discipline is Footprint vs Fingerprint. The strategic essay on why building editorial record is the work that compounds is The Editorial Record as the Most Valuable SEO Investment. The deeper-layer framework explaining how repeated Selection events compound into persistent visibility advantage is Retrieval Gravity.
For practitioners considering similar paid-placement programmes: run the five-question test from the Editorial Selection guide on each opportunity before committing budget. The test takes three minutes; the cumulative effect across an annual budget is the difference between a programme of one-off transactions and a programme that builds the compounding asset. The discipline is the routine that catches the moment of temptation before the spend is made. I run the routine now. I did not run it in January 2024. This report is what the absence of the routine produced.
