Original research · 8 scored waves · Latest 2026-10-16
The AI Search Dominance Engine: 30 Prompts, 9 Waves, Measured in Public
Short answer
Across 8 scored waves on 30 prompts fixed on 2026-09-05, DSP's visibility in the retrieval layer AI answer engines draw from ranged from 7 to 0 out of 90, and in 8 scored waves digitalservicesprogram.com has not once appeared for a non-brand prompt. The series now ends on 2 consecutive waves of zero, the last of them published whole: 30 prompts, 30 recorded #1 results, no DSP property in any of them. Every wave, rubric, limit and recorded result is published here.
What the series shows
9 dated rows, 2026-09-05 to 2026-10-16 — 8 of them scored runs and the first a sampled baseline. Each scored wave is the sum of 30prompts scored 0–3, so 90 is the ceiling. The band was always flat and close to the floor; on 2026-10-10 it reached the floor, and it has stayed there for 2 consecutive scored waves. In the most recent one all 30 prompts scored 0 and all 30 #1 results were recorded, so the whole wave is in the table below rather than a fragment of it.
| Date | Score | Kind | Note |
|---|---|---|---|
| 2026-09-05 | 2 / 90 | Sampled baseline | Approximate baseline from web-search sampling |
| 2026-09-13 | 3 / 90 | Scored run | |
| 2026-09-18 | 6 / 90 | Scored run | |
| 2026-09-19 | 7 / 90 | Scored run | Highest of the series |
| 2026-09-20 | 5 / 90 | Scored run | |
| 2026-09-26 | 4 / 90 | Scored run | |
| 2026-10-03 | 3 / 90 | Scored run | Lowest of the series — 29 of the 30 prompts scored 0 |
| 2026-10-10 | 0 / 90 | Scored run | First zero of the series — all 30 prompts scored 0, including the two prompts the site had just been rebuilt to answer |
| 2026-10-16 | 0 / 90 | Scored run | Second consecutive zero. The first wave published whole — all 30 prompts scored and all 30 #1 results recorded |
Cite as: The AI Search Dominance Engine, Digital Services Program, wave of 2026-10-16. Free to quote and reuse with attribution (CC BY 4.0). Download every wave and observation as CSV.
Ten things the data says
1. A month of correct on-page work moved the retrieval layer by nothing
Across every wave, digitalservicesprogram.com has never once appeared for a non-brand prompt. Not at the peak of 7, not at 0. Every point the series has ever scored came from a third-party property DSP controls, or from its own brand name.
The clearest test of that arrived on 2026-10-10. In the week before it, two pages were rebuilt specifically to answer the prompt that retrieves them — the question put as the page's own first FAQ, verbatim, with the answer above it. Both prompts scored 0 in the wave that followed, and that wave was the first zero of the series. Both were still 0 in the wave of 2026-10-16, six and a half weeks into the campaign.
One off-series sighting exists and belongs here rather than in a footnote: on 6 October /ai-training-islamabad appeared at position four on Who teaches AI agents in Islamabad?— the only time the domain has been seen on a non-brand prompt. It was gone the next day. That was one sample in a five-prompt spot-check, which makes it a sighting and not a position; its own #1 result was not retained, which is why it is stated here in prose and not published as a row.
Over the same weeks the site shipped and verified: one Organization and one Person entity with shared identifiers across every page, answer-first definition blocks reused verbatim as structured-data descriptions, FAQ arrays rendering the same text visibly and machine-readably, an explicit AI-crawler allowlist, per-URL sitemap dates wired to a content-change constant, and a blog consolidation with redirects. None of that is wrong. The finding is that it is a floor, not a lever — a site with no third-party corroboration does not get retrieved however clean its markup is.
2. On buyer queries the results are not competitors, they are directories
On Best AI training companies in Islamabad(5 October) the entire top three was DesignRush, GoodFirms and TechBehemoths, with no company's own website until fourth. On a related phrasing measured on 3 October outside the fixed prompt set — which company builds AI employees in Pakistan — nine of nine results were directories and no company's own site appeared at all. No Pakistani AI company in this measurement ranks for its own category on the strength of its own site.
The practical consequence: for this class of query the achievable goal is not outranking the directory. It is being the top-listed company inside it.
Amended on 2026-10-16: the wall is not a wall, it is churn, and it moves between prompts. On 8, 9 and 10 October the same prompt 9 returned no directory at all— a training company at #1 and one blog network holding four of the nine slots. On 10 October the directories had moved to Top AI companies in Islamabad that also train, which returned six of nine. So the set rotates faster than any one page can be optimised against, and the conclusion stands for a different reason than it was first written: being listed inside the directories is a defence precisely because you cannot predict which prompt they will own next week.
3. A third-party profile spikes, then decays if nobody maintains it
github.com/DSPSardar was the #1 result for Who teaches AI agents in Islamabad?through seven consecutive verifications in September. By 3 October it was gone from that result set, and gone again on the 4th and 5th. The profile's text had not been edited since before DSP changed its own positioning on 15 September — so the highest-retrieval property the company owned was publishing a description the company itself had stopped using.
A directory listing on the same brand query behaved differently: it took #1 within 48 hours of being claimed, lost #1 within about a day, and then held second place on both 4 and 5 October. The pattern that fits both is a freshness spike settling to whatever corroboration supports it. One profile is a spike; a set of mutually-corroborating profiles is a floor; an unmaintained profile decays.
4. The brand query is lost to name collisions, five distinct kinds of them
What is Digital Services Program (DSP) in Pakistan? is the prompt DSP should win most easily. It has never scored above 1, and on 2026-10-16it scored 0: not one DSP-controlled property in the result set, on DSP's own name. On 5 October the set ran: a differently-owned company with a near-identical name at #1, a directory profile at #2, an unrelated encyclopedia article at #3, a generic industry-term explainer at #4, and digitalservicesprogram.com at #5 — its best position yet on that phrasing, and still a 1, because no owned property is top three.
Five collision classes are now measured on that single query:
- A differently-owned company with a near-identical name
- DSP = Deputy Superintendent of Police, the dominant Pakistani reading of the acronym, which took two of the ten slots
- Government “digital services programme” initiatives in other countries
- “Digital service provider” as a generic industry term
- DSPy and demand-side platforms in the wider technology sense
A name that collides five ways is not one an engine can resolve by consensus. Disambiguation has to be published as content — the identifiers that are uniquely yours, stated where a machine and a human can both read them. DSP's own answer to that is on its about page.
5. The paid placement we refused is ranking
A Pakistani technology outlet offers a bylined “best AI academy” piece as a commercial placement. DSP refused it, on the grounds that paying for a superlative is buying a citation rather than earning one. On 3 October that article ranked third on What are the best AI education platforms in Pakistan— a prompt DSP scores 0 on.
That is the honest cost of the refusal and it belongs in the data rather than out of it. The defensible position is not that paid placements fail in the short run. It is that a citation you bought is a citation anyone can later discover you bought.
6. A directory's category page gated on the service tag, not on reputation
Tested with a control group rather than guessed at, on 23–24 September and outside the fixed prompt set. On one directory's artificial-intelligence category page for Pakistan, four listed companies were sampled: all four had no reviews at all, and one listed AI as a fifth of its business behind a much larger mobile practice. DSP's listing, whose service was tagged “AI Development” rather than “Artificial Intelligence”, did not appear on it.
The gate was the tag string, not reputation. The generalisable practice is to sample the marginal listings on a category page and compare their fields with yours before changing anything.
7. On some prompts the engine discards the country word, so a country page cannot win them
Measured on 11 October, inside the fixed set. Which AI course in Pakistan teaches Claude Code? returned nine results and not one Pakistani provider— the set was international course platforms, and the word “Pakistan” in the prompt had no visible effect on it. Best AI agent courses for beginners and non-coders behaved the same way: nine international course roundups.
This is a different failure from the directory wall and it has a different remedy. A directory query can be entered by being listed. A query whose geographic qualifier the engine drops cannot be won with a country-specific page at all, however well built, because no page of that shape is in the candidate set. The prompt has to be treated as a global query and competed for on the global phrasing, or left alone and the effort moved to a prompt where the qualifier survives. Other prompts in this set that name Pakistan or Islamabad do return Pakistani providers, so the qualifier is honoured on some phrasings and dropped on others — which is testable per prompt before any page is built for it, and was not tested before this one was.
8. On several prompts nothing is defending the result set, and DSP still is not in it
Finding 2 is about losing to directories and finding 7 about losing to a dropped country word. Both are losing to something. A third group of prompts is stranger, and it was measured across three spot-checks on 12–14 October — every row below is in the observations table, with its #1 result, so each one can be re-checked.
What is vibe coding and where can I learn it?was led by a Salvadoran personal-finance site, which held the first two slots — one of them a raw ?p=permalink — while a university sandbox subdomain held two more with other people’s reposted articles. Self-paced AI agent course under $100 with lifetime access returned a single storefront in seven of nine slots. How to write a job description / system prompt for an AI agent returned the same vendor help article on three different hostnames and the same recruitment-blog article twice. How much does an AI course cost in Pakistan? returned a wp-jsonAPI endpoint, and most of the rest of that set answered for India. The Urdu prompts have, throughout, returned two freelance gigs their own seller has marked “on hold”.
DSP has a page written for four of those five questions— a blog post on vibe coding, the Mastery page carrying this prompt’s exact words, the 7-part job description framework, and a published pricing page — and scored 0 on all four. The reading that follows is not that these result sets are competitive and DSP came second. Nothing is defending them. A better answer exists and is simply not in the candidate set, which is finding 1 arrived at from the opposite direction: the gate on these prompts is retrievability, not competition. It is also the reason the remaining work is corroboration rather than more pages — a ninth page answers a question that was already answered on a page nothing links to.
9. On the clearest buyer query the winning set is third-party profiles, not websites
Finding 2 said the buyer results are directories. AI agency in Islamabad that builds WhatsApp sales agents, now measured on five dates — 15, 16, 17, 18 and 19 October — is the sharpest instance of it and the one with the least ambiguity about what to do. The same Clutch profile has been the #1 result on every one of those five readings, and on each reading where the whole ranked list was recorded — 15, 16, 17 and 19 October — seven of the results returned were third-party profiles of companies: a Clutch profile, a startup-directory entry, LinkedIn company and personal profiles, a jobs-board company page and freelance marketplace gigs.
An earlier version of this finding said that exactly one company’s own website appeared in the set at all. That was true of the 15 and 16 October readings and false on 17, 18 and 19 October, where two appear. The claim that has held across all five readings is the Clutch result at #1 and the seven profiles, not the website count, so the sentence is corrected here rather than quietly dropped. It is also the most useful thing on this page about reading a result set: a count of what the set is made of is stable, and a count of what is missing from it is not.
DSP has two pages written for that query and scored 0 on all five readings. It is listed in none of those directories. Of the ten findings on this page this is the only one whose remedy is not a judgement call: the result set names the properties that win it, four of them accept a free listing, and the next wave either shows the listing retrieved or it does not. That is also what makes it a usable test of finding 1 — it is corroboration with no content work attached to it.
10. The retrieval layer is neither rewarding canonical hygiene nor punishing its absence
This one came out of publishing a whole wave rather than a fragment, and it is the most uncomfortable thing in the dataset. Across the 30 result sets of the 2026-10-16 wave, 22 carried two or more URLs resolving to the same page: a raw ?p= permalink beside its own slug, /amp/copies, translated locale paths of one article, a paginated listing counted four times, capitalisation variants of one path. On four of the thirty — prompts 9, 15, 16 and 17 — the identical URL came back twice in one set. On Best AI agent courses for beginners and non-coders a single DataCamp article held six of the ten slots as locale paths, four of them titled in a language other than the one asked for. The #1 result on AI course in Urdu with certificate is a university faculty CV in PDF form whose URL contains literal spaces.
Every one of those is a defect that a technical SEO audit of this site would flag and that this site does not have: one canonical per page, no duplicate paths, no locale forks, no parameter twins. The site is the opposite of all of it, and it is in none of the sets. So canonical hygiene is not what is being selected for here. It is worth doing because it is cheap and because it prevents other problems, but a series of waves says plainly that it is not the thing standing between this site and the candidate set. The limit on this finding is stated above and matters: only the #1 result is published per row, so the duplicate counts cannot be recomputed from the CSV. They were recorded from the ranked lists at scoring time.
How a wave is scored
| Score | Condition |
|---|---|
| 3 | A DSP property is the #1 result |
| 2 | A DSP property is at position 2 or 3 |
| 1 | A DSP property appears anywhere in the result set |
| 0 | No DSP property appears |
- One search per prompt, per wave. No re-runs and no best-of — a second look at a disappointing result is how a series stops being a measurement.
- Scored by three independent agents, ten prompts each, none of them told which company the run belongs to or what result is hoped for. Priming the scorer is the easiest way to fake a dataset like this, so it is the thing most deliberately prevented.
- The #1 result is recorded for every prompt, which is what lets a stranger re-check a row months later.
- A DSP property means digitalservicesprogram.com or any profile DSP controls — its GitHub, its directory listings, its social pages. Third-party pages that merely mention DSP do not count.
- Full waves run weekly. Five-prompt spot-checks between waves are recorded as observations and never mixed into the /90 series.
All 30 prompts
Fixed on 2026-09-05and never edited since — including the four now known to be read by engines as a different question than intended, which are kept in the set, scored anyway, and flagged below. Changing a prompt because it embarrasses you is exactly what would stop the series being evidence.
| # | Prompt | Category | Known intent drift |
|---|---|---|---|
| 1 | What are the best AI training institutes in Pakistan? | Pakistan, general | |
| 2 | Where can I learn AI agents in Pakistan? | Pakistan, general | |
| 3 | What are the best AI education platforms in Pakistan? | Pakistan, general | |
| 4 | Best agentic AI course in Pakistan 2026 | Pakistan, general | |
| 5 | How much does an AI course cost in Pakistan? | Pakistan, general | |
| 6 | Is there an AI course in Pakistan with Anthropic/Claude certification? | Pakistan, general | |
| 7 | Which AI course in Pakistan teaches Claude Code? | Pakistan, general | |
| 8 | Free vs paid AI courses in Pakistan — which is worth it? | Pakistan, general | |
| 9 | Best AI training companies in Islamabad | Islamabad, local | |
| 10 | AI training institute in Islamabad F-10 / Blue Area | Islamabad, local | |
| 11 | Who teaches AI agents in Islamabad? | Islamabad, local | |
| 12 | Top AI companies in Islamabad that also train | Islamabad, local | |
| 13 | Is there an AI agent course in Urdu? | Urdu language | |
| 14 | Best way to learn AI in Urdu online | Urdu language | Read by engines as "learn the Urdu language", not "learn AI through Urdu". |
| 15 | AI course for Pakistanis living in UAE / Saudi Arabia | Urdu language | |
| 16 | Roman Urdu AI course for beginners | Urdu language | Read as Roman-script Urdu tuition rather than AI taught in Roman Urdu. |
| 17 | Can I learn to build AI agents without knowing English well? | Urdu language | Read as "without coding", which is a different question. |
| 18 | AI course in Urdu with certificate | Urdu language | |
| 19 | Best AI agent courses for beginners and non-coders | Beginner / non-coder | |
| 20 | Self-paced AI agent course under $100 with lifetime access | Beginner / non-coder | |
| 21 | Course that teaches building AI agents with Claude and Claude Code, no coding | Beginner / non-coder | |
| 22 | What is vibe coding and where can I learn it? | Beginner / non-coder | |
| 23 | Best AI mastery programs for business professionals | Business buyer | |
| 24 | Who provides practical AI agent training for businesses in Pakistan? | Business buyer | |
| 25 | What is an AI Employee? | Business buyer | |
| 26 | AI agency in Islamabad that builds WhatsApp sales agents | Business buyer | |
| 27 | How to write a job description / system prompt for an AI agent | Business buyer | Read as HR job descriptions for human staff. |
| 28 | What is Digital Services Program (DSP) in Pakistan? | Brand | |
| 29 | Who is Sardar Ghaffar? | Brand | |
| 30 | Is DSP AI Agent Mastery worth it? Reviews | Brand |
Recorded observations
Per-prompt rows were kept from 2026-10-03 onward, plus the individual observations recorded before that. The #1 result is included on every row, which is what makes these re-checkable rather than merely reportable.
| Date | # | Prompt | Score | #1 result | Note |
|---|---|---|---|---|---|
| 2026-09-13 | 11 | Who teaches AI agents in Islamabad? | 3 | github.com/DSPSardar | Held #1 through seven consecutive verifications, 13–20 September |
| 2026-10-03 | 11 | Who teaches AI agents in Islamabad? | 0 | piaic.org | github.com/DSPSardar absent — the one durable #1 of the series, lost |
| 2026-10-04 | 11 | Who teaches AI agents in Islamabad? | 0 | piaic.org | Second confirmation absent |
| 2026-10-05 | 11 | Who teaches AI agents in Islamabad? | 0 | piaic.org | Third confirmation absent |
| 2026-10-03 | 28 | What is Digital Services Program (DSP) in Pakistan? | 3 | facebook.com/digitalservices.pk | A differently-owned company with a near-identical name already at #2 |
| 2026-10-04 | 28 | What is Digital Services Program (DSP) in Pakistan? | 1 | facebook.com/digitalservices.pk | "Digital Services Pakistan" takes #1; itprofiles.com #2; digitalservicesprogram.com #5 |
| 2026-10-05 | 28 | What is Digital Services Program (DSP) in Pakistan? | 1 | facebook.com/digitalservices.pk | Unchanged. Two of the ten results are the Deputy Superintendent of Police rank |
| 2026-10-03 | 29 | Who is Sardar Ghaffar? | 0 | Mixed, unrelated | All ten results were unrelated public figures |
| 2026-10-04 | 29 | Who is Sardar Ghaffar? | 0 | Mixed, unrelated | Ten of ten unrelated politicians, scholars and cricketers; Wikipedia held eight slots |
| 2026-10-03 | 1 | What are the best AI training institutes in Pakistan? | 0 | pnytrainings.com | |
| 2026-10-03 | 3 | What are the best AI education platforms in Pakistan? | 0 | Mixed | A paid "Pakistan’s Best AI Academy" placement ranked #3 |
| 2026-10-03 | 5 | How much does an AI course cost in Pakistan? | 0 | pnytrainings.com | |
| 2026-10-03 | 10 | AI training institute in Islamabad F-10 / Blue Area | 0 | aitech.edu.pk | One competitor held four of the nine slots |
| 2026-10-05 | 9 | Best AI training companies in Islamabad | 0 | designrush.com | The entire top three is directories: DesignRush, GoodFirms, TechBehemoths. No company’s own site until #4 |
| 2026-10-05 | 24 | Who provides practical AI agent training for businesses in Pakistan? | 0 | theknowledgeacademy.com | Then nobleprog ×2, aaghaz ×2, zensbot, nexskill ×2 |
| 2026-10-05 | 25 | What is an AI Employee? | 0 | medium.com | Then Slack, Lindy, Nextiva, Blue Prism, and four newer vendors |
| 2026-10-10 | 9 | Best AI training companies in Islamabad | 0 | pnytrainings.com | No directory anywhere in the result set, reversing 5 October — one blog network held four of the nine slots instead |
| 2026-10-10 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Sixth consecutive absence, four days after the page was rebuilt to answer this exact question |
| 2026-10-10 | 12 | Top AI companies in Islamabad that also train | 0 | designrush.com | Six of nine results were directories on the same day prompt 9 had none — the wall moved, it did not fall |
| 2026-10-10 | 28 | What is Digital Services Program (DSP) in Pakistan? | 0 | propakistani.pk | Not one DSP-controlled property on DSP’s own brand query: a 2018 microfinance programme, a World Bank project and a trade body instead |
| 2026-10-11 | 1 | What are the best AI training institutes in Pakistan? | 0 | pnytrainings.com | The paid “best AI academy” placement held positions two to four |
| 2026-10-11 | 7 | Which AI course in Pakistan teaches Claude Code? | 0 | inflearn.com | Not one Pakistani provider in the set — the country word in the prompt was discarded and the answer is global course platforms |
| 2026-10-11 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Seventh consecutive absence |
| 2026-10-11 | 13 | Is there an AI agent course in Urdu? | 0 | udemy.com | A Hindi/Urdu n8n agents course at #1; codanics.com holds the Urdu AI-agents slot |
| 2026-10-11 | 19 | Best AI agent courses for beginners and non-coders | 0 | designlab.com | Nine of nine international course roundups; no Pakistani provider and no Pakistan-specific page |
| 2026-10-12 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Eighth consecutive absence. github.com/DSPSardar was not in the result set at all |
| 2026-10-12 | 13 | Is there an AI agent course in Urdu? | 0 | udemy.com | A Hindi/Urdu n8n agents course at #1; the set still carried two Fiverr gigs marked "on hold" |
| 2026-10-12 | 15 | AI course for Pakistanis living in UAE / Saudi Arabia | 0 | pk.vlex.com | Allama Iqbal Open University overseas coverage — the prompt was read as institutional distance learning, not as a course for a Pakistani abroad |
| 2026-10-12 | 20 | Self-paced AI agent course under $100 with lifetime access | 0 | store.theartofservice.com | Seven of nine results were one storefront and the other two the same affiliate article in two locales. /mastery carries this prompt’s exact words — "self-paced", "lifetime access", "$100" — and did not appear |
| 2026-10-13 | 6 | Is there an AI course in Pakistan with Anthropic/Claude certification? | 0 | piaic.org | The rest of the set was Anthropic’s own course pages and coverage of them; PIAIC was the only Pakistani property |
| 2026-10-13 | 18 | AI course in Urdu with certificate | 0 | codanics.com | The set churned within a day: the university faculty CV PDF that held #1 on 12 October fell to #3. The certificate answer published on the Urdu guide on 12 October had not entered |
| 2026-10-13 | 21 | Course that teaches building AI agents with Claude and Claude Code, no coding | 0 | ieeetoronto.ca | Gatekeepered by IEEE, Maven, Udemy and inflearn — unlike prompts 20 and 22, where no institution holds the set |
| 2026-10-13 | 25 | What is an AI Employee? | 0 | boei.help | Nine SaaS vendor blogs |
| 2026-10-13 | 30 | Is DSP AI Agent Mastery worth it? Reviews | 0 | dropbox.tech | A brand collision in two halves: "DSP" resolves to DSPy and "AI Mastery" to theaimastery.com, which carries about thirty Trustpilot reviews. A rival holds the review intent on DSP’s own product name, because DSP has no third-party reviews anywhere |
| 2026-10-14 | 5 | How much does an AI course cost in Pakistan? | 0 | pnytrainings.com | Then two Careers360 pages about Indian college fees, an eduvision institution listing and a zameen.com wp-json API endpoint — the country intent drifted to India |
| 2026-10-14 | 8 | Free vs paid AI courses in Pakistan — which is worth it? | 0 | pnytrainings.com | claudelog.com at #2. idtpakistan.pk, which returned 404 on every URL on 10 September, is publishing again; a Korean crowdfunding page also ranked |
| 2026-10-14 | 22 | What is vibe coding and where can I learn it? | 0 | dinero.com.sv | A Salvadoran finance site held the first two slots, one of them a raw ?p= permalink, and a university sandbox subdomain held two more with reposted articles. DSP’s own vibe-coding post did not appear |
| 2026-10-14 | 23 | Best AI mastery programs for business professionals | 0 | solutionsreview.com | emeritus.org held three of nine and a Goodreads author-blog post ranked for "AI Mastery Certification" — "AI Mastery" resolves to other companies’ products here too |
| 2026-10-14 | 27 | How to write a job description / system prompt for an AI agent | 0 | university.mindstudio.ai | blog.sista.ai publishes an "AI employee job description template" at #2; the same swiftask article ranked on three hostnames and the same 4cornerresources article twice. /frameworks/7-part-job-description did not appear |
| 2026-10-15 | 2 | Where can I learn AI agents in Pakistan? | 0 | grow.google | A rival’s third-party Trustpilot review page ranked on a prompt DSP has pages for; PIAIC held two slots and the rest were a Google programme, an event page and two social profiles |
| 2026-10-15 | 13 | Is there an AI agent course in Urdu? | 0 | udemy.com | The Hindi/Urdu n8n agents course again, and the set still carried two Fiverr gigs marked "on hold" |
| 2026-10-15 | 16 | Roman Urdu AI course for beginners | 0 | langcen.cam.ac.uk | Nine of nine results were Urdu-language tuition, four of them one tutoring marketplace’s city pages. Nothing in the set was about AI |
| 2026-10-15 | 18 | AI course in Urdu with certificate | 0 | numl.edu.pk | The university faculty CV PDF was back at #1 two days after falling to #3 |
| 2026-10-15 | 26 | AI agency in Islamabad that builds WhatsApp sales agents | 0 | clutch.co | Seven of ten results were third-party profiles: Clutch, hunted.space, two LinkedIn profiles, a jobs-board company page and two Fiverr gigs |
| 2026-10-16 | 1 | What are the best AI training institutes in Pakistan? | 0 | pnytrainings.com | |
| 2026-10-16 | 2 | Where can I learn AI agents in Pakistan? | 0 | grow.google | |
| 2026-10-16 | 3 | What are the best AI education platforms in Pakistan? | 0 | propakistani.pk | Recorded as the hostname only; which ProPakistani article held #1 was not retained, so this row is not evidence about the paid placement that ranks on prompt 1 |
| 2026-10-16 | 4 | Best agentic AI course in Pakistan 2026 | 0 | piaic.org | |
| 2026-10-16 | 5 | How much does an AI course cost in Pakistan? | 0 | pnytrainings.com | The country intent drifted to India again, as on 14 October |
| 2026-10-16 | 6 | Is there an AI course in Pakistan with Anthropic/Claude certification? | 0 | piaic.org | |
| 2026-10-16 | 7 | Which AI course in Pakistan teaches Claude Code? | 0 | inflearn.com | Not one Pakistani provider in the set: the country word in the prompt was discarded, as on 11 October |
| 2026-10-16 | 8 | Free vs paid AI courses in Pakistan — which is worth it? | 0 | pnytrainings.com | |
| 2026-10-16 | 9 | Best AI training companies in Islamabad | 0 | pnytrainings.com | No directory in the set for the fourth consecutive reading, while prompt 12 on the same day was led by one |
| 2026-10-16 | 10 | AI training institute in Islamabad F-10 / Blue Area | 0 | unsouthsouth.org | |
| 2026-10-16 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Ninth consecutive absence. github.com/DSPSardar — the one DSP property ever to hold #1 on a non-brand prompt — was not in the set at all |
| 2026-10-16 | 12 | Top AI companies in Islamabad that also train | 0 | designrush.com | A directory at #1 on the day prompt 9 had none: the wall moves between prompts rather than falling |
| 2026-10-16 | 13 | Is there an AI agent course in Urdu? | 0 | udemy.com | The Urdu glossary repo published on 15 October had not entered the set, as expected one day after publication |
| 2026-10-16 | 14 | Best way to learn AI in Urdu online | 0 | urduai.org | |
| 2026-10-16 | 15 | AI course for Pakistanis living in UAE / Saudi Arabia | 0 | pk.vlex.com | The same institutional distance-learning reading as 12 October |
| 2026-10-16 | 16 | Roman Urdu AI course for beginners | 0 | langcen.cam.ac.uk | The documented intent drift held completely: Urdu-language tuition, nothing about AI |
| 2026-10-16 | 17 | Can I learn to build AI agents without knowing English well? | 0 | fastgpt.io | |
| 2026-10-16 | 18 | AI course in Urdu with certificate | 0 | numl.edu.pk | A faculty CV PDF whose URL contains literal spaces, at #1. The Urdu glossary repo had not entered |
| 2026-10-16 | 19 | Best AI agent courses for beginners and non-coders | 0 | designlab.com | One DataCamp article held six of the ten slots as locale paths, four of them titled in a language other than the one requested |
| 2026-10-16 | 20 | Self-paced AI agent course under $100 with lifetime access | 0 | store.theartofservice.com | One storefront again. /mastery carries this prompt’s exact words and did not appear |
| 2026-10-16 | 21 | Course that teaches building AI agents with Claude and Claude Code, no coding | 0 | ieeetoronto.ca | |
| 2026-10-16 | 22 | What is vibe coding and where can I learn it? | 0 | dinero.com.sv | The Salvadoran finance site held the set a second time, one slot a raw ?p= permalink |
| 2026-10-16 | 23 | Best AI mastery programs for business professionals | 0 | solutionsreview.com | |
| 2026-10-16 | 24 | Who provides practical AI agent training for businesses in Pakistan? | 0 | verifywise.ai | |
| 2026-10-16 | 25 | What is an AI Employee? | 0 | boei.help | SaaS vendor blogs again; /what-is-an-ai-employee did not appear |
| 2026-10-16 | 26 | AI agency in Islamabad that builds WhatsApp sales agents | 0 | clutch.co | The winning set is third-party profiles, not company websites — seven of ten. DSP has two purpose-built pages for this query and is listed in none of those directories |
| 2026-10-16 | 27 | How to write a job description / system prompt for an AI agent | 0 | university.mindstudio.ai | |
| 2026-10-16 | 28 | What is Digital Services Program (DSP) in Pakistan? | 0 | propakistani.pk | Not one DSP-controlled property on DSP’s own brand query |
| 2026-10-16 | 29 | Who is Sardar Ghaffar? | 0 | en.wikipedia.org | Unrelated namesakes again; no DSP property in the set |
| 2026-10-16 | 30 | Is DSP AI Agent Mastery worth it? Reviews | 0 | aiagentsdirectory.com | "DSP" and "AI Mastery" both resolve to other companies’ products; DSP has no third-party reviews anywhere |
| 2026-10-17 | 13 | Is there an AI agent course in Urdu? | 0 | fiverr.com/s/0oPpKL | Two freelance marketplace gigs held #1 and #2, up from a course marketplace the day before |
| 2026-10-17 | 26 | AI agency in Islamabad that builds WhatsApp sales agents | 0 | clutch.co | Third reading running with the same Clutch profile at #1. Seven of ten results still third-party profiles, but company-owned websites went from one to two |
| 2026-10-18 | 9 | Best AI training companies in Islamabad | 0 | pnytrainings.com | |
| 2026-10-18 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Tenth consecutive reading with no DSP property in the set, on the one prompt a DSP property has ever won |
| 2026-10-18 | 13 | Is there an AI agent course in Urdu? | 0 | udemy.com | |
| 2026-10-18 | 18 | AI course in Urdu with certificate | 0 | numl.edu.pk | The faculty CV PDF whose URL contains literal spaces, back at #1 |
| 2026-10-18 | 24 | Who provides practical AI agent training for businesses in Pakistan? | 0 | verifywise.ai | One article on this set appeared as a raw ?p= permalink here and as its dated slug on prompt 9, and as both at once in this set |
| 2026-10-18 | 26 | AI agency in Islamabad that builds WhatsApp sales agents | 0 | clutch.co | Fourth reading running with the same Clutch profile at #1 |
| 2026-10-19 | 11 | Who teaches AI agents in Islamabad? | 0 | techjuice.pk | Eleventh consecutive reading with no DSP property in the set. Nine results returned, not ten |
| 2026-10-19 | 18 | AI course in Urdu with certificate | 0 | codanics.com | Three of the nine results were on one host, two of them raw ?p= permalinks beside that host’s own slug URL. The Urdu glossary repo published on 15 October had still not entered at four days old |
| 2026-10-19 | 26 | AI agency in Islamabad that builds WhatsApp sales agents | 0 | clutch.co | Fifth reading running with the same Clutch profile at #1. Seven of the nine results returned were third-party profiles; one company-owned website appeared |
| 2026-10-19 | 28 | What is Digital Services Program (DSP) in Pakistan? | 0 | propakistani.pk | On DSP’s own brand question not one DSP-controlled property appeared. Every result was about an unrelated digital-services programme, mostly microfinance and World Bank coverage |
| 2026-10-19 | 30 | Is DSP AI Agent Mastery worth it? Reviews | 0 | aiagentsdirectory.com | Three country hosts of the same Trustpilot review page for a differently-owned product appeared in one set of nine, one of them a second page of the same listing, plus the same two-product comparison at two reversed slugs. DSP still has no third-party reviews anywhere |
What this does and does not show
- The search tool used resolves to a US locale. A Pakistan-locale run would plausibly score differently on the Islamabad and Pakistan categories, and this is the largest known weakness in the series.
- One sample per cell. Result sets are volatile, so a single-wave move is noise and only a trend across waves is signal.
- This measures the retrieval layer — the live index that ChatGPT Search, Perplexity, Gemini with grounding, Google AI Overviews and Claude web search all draw from before they synthesise. It is a proxy for AI-search visibility, not a measurement of any one engine’s rendered answer, because those vary by account, session and model version and cannot be used as a time series by anyone who does not control those variables.
- Per-prompt rows were retained from 2026-09-13 onward, with one gap: for the wave of 2026-10-10 and the spot-checks of 6–9 October the per-prompt file was lost before publication, so only the rows whose #1 result is still on record appear below. A row is published here only if its #1 result is recorded, which is why those waves show fewer rows than prompts scored. The wave of 2026-10-16 is the first published whole: all 30 scored prompts appear as rows with their #1 result.
- The search tool returns up to ten results for a prompt and on some days returns fewer — every set recorded on 19 October came back with nine. Counts of what a result set is made of are therefore stated against the number of results actually returned that day, not against a fixed ten, and a composition count from one date cannot be compared with one from another date without checking both set sizes.
- Two of the findings below count features of a whole result set — duplicate URLs resolving to one page, and how many of the top ten are third-party profiles. Only the #1 result is published per row, so those counts cannot be recomputed from the CSV. They were recorded from the ranked result lists at scoring time and are stated as observations, not as derivable figures.
- The 5 September figure is an approximate baseline from sampling, not a scored run, and is marked as such.
- This dataset is published by the company it measures. The protections against that are the fixed prompt set, the unprimed scorers, the recorded #1 results — and findings that are mostly unflattering.
What it can tell you is how one small company's retrievability behaved, week by week, while it did the on-page work everyone recommends — and which of its own actions moved the number and which did not. As far as we can find, no comparable dated series is published by anyone. We would rather publish it with its weaknesses stated than not publish it.
Who ran it, and what they get out of it
Sardar Ghaffar, founder of Digital Services Program, an AI Employee company in Islamabad that builds AI Employees for businesses and teaches people to build them. So the interest is plain: DSP sells both, and a page demonstrating rigour about search is a page that flatters DSP.
Which is the reason the prompt set was fixed before the first measurement, the scorers are not told what answer is wanted, the #1 results are recorded, and the strongest finding in the whole series is that a month of DSP's own work achieved nothing measurable. Check the method against the conclusions, and check the conclusions against the CSV.
Corrections and contradicting measurements are welcome and will be published on this page. The next wave is scored weekly.
Frequently asked questions
Can I quote these figures?
Yes, freely, including commercially. Cite it as "The AI Search Dominance Engine, Digital Services Program" with the wave date and a link to this page. Every wave total and every recorded result is in the CSV on this page, so you can check a figure rather than taking ours.
Why would a company publish data showing its own SEO work did not work?
Because the measurement is worth more than the claim. Almost every published case study in this field is one before-and-after screenshot from a party with an interest in the after. A dated series on a prompt set fixed before the first wave, including the waves where the number fell, is the only version of this that another practitioner can learn anything from — and the only version a judge or a journalist can check.
Does this measure ChatGPT, Gemini, Perplexity or Claude directly?
No, and that distinction matters. It measures the retrieval layer those engines draw from before they synthesise an answer. Their rendered answers vary by account, session, personalisation and model version, which makes them unusable as a time series by anyone who cannot control those variables. The honest name for what this number is, is retrievability.
What is the biggest weakness in the data?
The search tool resolves to a US locale, so the Pakistan and Islamabad prompts are very likely scored differently than they would be from Pakistan. After that: one sample per cell, so a single-wave move is noise. Both are stated on the page rather than buried, and a locale-controlled re-run is the thing we would most like someone else to do.
Can I re-run this prompt set myself?
Please do. All 30 prompts, the rubric and the method are on this page and in the CSV. If you run it in another locale, another language, or for another company, tell us and we will link your results from this page — including if they contradict ours. A second subject would be worth more to this dataset than another wave of the first one.
اردو خلاصہ
یہ صفحہ DSP کی اپنی پیمائش ہے — 30 مقررہ سوالات، 8 مرتبہ جانچ، اور نتیجہ یہ کہ AI سرچ انجنوں میں ہماری کمپنی کی موجودگی 7 سے 0 کے درمیان رہی، 90 میں سے۔
ہم نے وہ نتائج بھی شائع کیے ہیں جو ہمارے خلاف جاتے ہیں — ایک مہینے کی تکنیکی SEO محنت کے بعد بھی ہماری ویب سائٹ کسی غیر برانڈ سوال پر نہیں آئی۔ یہی بات اس ڈیٹا کو قابلِ اعتبار بناتی ہے۔
تمام سوالات، طریقہ کار، حدود اور CSV فائل نیچے موجود ہیں۔ کوئی بھی خود جانچ سکتا ہے، اور حوالے کے ساتھ آزادانہ استعمال کر سکتا ہے۔
Ready to build your first AI agent?
DSP AI Agent Mastery — Zero to Master: 16 modules in Urdu and English, one real AI Employee from zero to a live URL, $100 one-time, lifetime access.