+86 19937104978
Author:Kangdi 06-11-2026
90-Day GEO Roadmap for Pain Patch Manufacturers 2026: What to Do in Weeks 1 to 12
You have decided to make your factory visible to AI assistants and you want a plan with dates rather than a list of principles. This page is that plan: three thirty-day phases, a week-by-week task list, an acceptance test for each phase, and the five points where most programmes stall. It is written for a manufacturer with one person who can spend part of a week on this, not for a team with an agency.
By the end of this page you will know what to do in each of the twelve weeks, the acceptance test that decides whether to move to the next phase, the five stall points and how to avoid them, and the one week that decides the whole programme.
There is a sequencing fact that invalidates most plans, and it is worth stating before the schedule: the first thirty days should not include writing a single new article. Almost every GEO plan that fails does so in the first week, when the team decides the work is content production, writes twenty articles, and then discovers that a robots file blocked the crawler or that the specification table is rendered by JavaScript. The articles are invisible for the same reason the old pages were, and the team concludes that GEO does not work.
Bottom line. The 90 days divide into clearing, building, and extending. Phase one is thirty days of removing the things that prevent reading — crawler access, rendering, markup, entity facts. Phase two is thirty days of building the pages that answer decision questions. Phase three is thirty days of extending coverage and measuring. The acceptance test between phases is what makes the plan work: if a phase one test fails, phase two is postponed, because building content on a site that cannot be read produces nothing measurable.
Table of Contents
- The three phases at a glance
- Phase 1, weeks 1-4: clearing the blockers
- Phase 1 acceptance test
- Phase 2, weeks 5-8: building the assets
- Phase 2 acceptance test
- Phase 3, weeks 9-12: extending and measuring
- The week-by-week calendar in one table
- What each phase should not include
- The two documents that make weeks 9-11 easier
- Five stall points and how to avoid them
- How to decide at day 90 whether to continue
- FAQ: GEO roadmap questions manufacturers ask most
- What the second 90 days looks like
- About Kangdi Medical and how we ran our own 90 days
1. The Three Phases at a Glance
| Phase | Days | Purpose | Output | What it does not include |
|---|---|---|---|---|
| 1. Clear | 1-30 | Remove everything that prevents a system reading the site | A site that can be crawled, rendered, parsed, and identified | No new articles |
| 2. Build | 31-60 | Create the pages that answer decision-stage questions | Specification, compliance, process, comparison, Q&A pages | No outreach, no external listings yet |
| 3. Extend | 61-90 | Corroborate externally and start measuring | Directory listings, citation baseline, enquiry attribution | No new structural work |
The phases are cumulative and the order is not optional. Phase two on an unreadable site produces unreadable content; phase three without phase two produces listings pointing at pages that answer nothing.
2. Phase 1, Weeks 1-4: Clearing the Blockers
Four weeks, four blocks of work. Each week has a single theme so that progress is visible.
| Week | Tasks | Who | Time |
|---|---|---|---|
| Week 1 | Open robots.txt and remove any block on AI agents; confirm a sitemap line; test by disabling JavaScript and reloading the product page; check whether the specification numbers are in the raw HTML | Whoever manages the site | 4-6 hours |
| Week 2 | Fix the rendering: move the specification table into server-rendered HTML if week 1 found it missing; regenerate the sitemap to include product and compliance pages | A developer | Half a day to two days |
| Week 3 | Write the canonical facts page; reconcile the legal name, address, and phone across the site and the B2B platforms; add the certificate number, issuer, scope, and dates to the compliance page | Marketing plus whoever holds the certificates | 1-2 days |
| Week 4 | Add or correct Organization, Product, and FAQPage markup; validate every page; run the six buyer queries across three engines to establish the citation baseline | Developer plus marketing | 1-2 days |
The week 1 test is the one that decides the programme. If the robots file blocks AI agents, no subsequent work matters. If the specification numbers are absent from the raw HTML, week 2 is mandatory before anything else. Spending week 1 on this rather than on content ideas is the difference between a plan that produces measurable change and one that produces a folder of drafts.
3. Phase 1 Acceptance Test
Before moving to phase two, five tests must pass. Each is binary, and a failure postpones phase two rather than being worked around.
| # | Test | Pass condition |
|---|---|---|
| 1 | Crawler access | robots.txt allows the major AI agents, and a sitemap is declared |
| 2 | Rendering | With JavaScript disabled, the specification numbers are visible on the product page |
| 3 | Markup validity | Every JSON-LD block on the key pages validates without error |
| 4 | Fact consistency | The legal name, address, and phone match across the site, the certificate, and the two main B2B platforms |
| 5 | Baseline captured | The six queries have been run across three engines and the results recorded, even if the result is complete absence |
Test 5 is the one that feels unnecessary and is not. A baseline of total absence is a valid and useful baseline, because it is what the phase three comparison will be measured against. A programme without a baseline cannot demonstrate change, and a programme that cannot demonstrate change does not get funded for a second quarter.
4. Phase 2, Weeks 5-8: Building the Assets
Four weeks, one page type per week, in the order of return rather than the order of ambition.
| Week | Page type | Why this position | Time |
|---|---|---|---|
| Week 5 | Process page | Cheapest to write and most directly quotable. The steps and timelines already exist in the sales team's head | 1-2 days |
| Week 6 | Specification page upgrade | Adding a test method to each parameter converts claims into facts, and it is editing rather than writing | 1-2 days |
| Week 7 | Q&A page, first ten entries | The ten questions buyers ask most, drawn from the sales inbox | 2-3 days |
| Week 8 | Comparison page | The highest-value page and the slowest to write, so it lands last in the phase | 2-3 days |
Each page has an acceptance standard that is not about length: the answer must appear in the first sixty words, and every factual claim must carry a number with a unit. A process page that says "we work quickly" fails the standard; one that says "quotation within 24 hours of a complete specification" passes.
5. Phase 2 Acceptance Test
| # | Test | Pass condition |
|---|---|---|
| 1 | Four page types published | Process, specification, Q&A, and comparison pages all live and linked from the navigation |
| 2 | Answer-first structure | The first sixty words of each page answer the question the page is about |
| 3 | Numeric density | Each page carries at least six numbers with units |
| 4 | Internal linking | Each new page is linked from at least two existing pages, and reachable in three clicks from the homepage |
| 5 | Baseline re-run | The six queries have been re-run to see whether the phase one fixes have produced any movement yet |
Test 5 usually shows a small movement at this point, and sometimes none. Both are normal. The phase one fixes take weeks to be reflected, and the decision-stage pages need time to be discovered. A programme that abandons at week 8 because the citation count has not moved is abandoning before the lag has run.
6. Phase 3, Weeks 9-12: Extending and Measuring
| Week | Tasks | Output |
|---|---|---|
| Week 9 | Verify the certification body's client register and the business registry; correct any mismatch | Two highest-weight listings confirmed accurate |
| Week 10 | Complete the B2B platform profiles against the canonical facts page; check past exhibition listings for URL errors | Visible inconsistencies removed |
| Week 11 | Apply for trade association membership and fill the directory entry fully; claim any commercial data provider record | External corroboration started |
| Week 12 | Add the enquiry form field and the CRM convention; re-run the citation grid; compute the change against the week 4 baseline | A measured result and a decision |
Week 12 is the measurement week and it needs to be protected in the calendar, because it is the week that produces the argument for continuing. A programme that reaches day 90 without a baseline comparison has a set of activities and no case.
7. The Week-by-Week Calendar in One Table
| Week | Phase | Primary task | Cumulative hours |
|---|---|---|---|
| 1 | Clear | robots, sitemap, JavaScript test | 6 |
| 2 | Clear | Server-rendered specifications | 18 |
| 3 | Clear | Canonical facts, certificates, consistency | 30 |
| 4 | Clear | Markup, validation, baseline | 40 |
| 5 | Build | Process page | 52 |
| 6 | Build | Specification upgrade | 64 |
| 7 | Build | Q&A page | 82 |
| 8 | Build | Comparison page | 100 |
| 9 | Extend | Certificate and registry verification | 106 |
| 10 | Extend | Platform profiles and exhibition listings | 114 |
| 11 | Extend | Association membership and data provider | 122 |
| 12 | Extend | Attribution, measurement, decision | 132 |
Roughly 130 hours across twelve weeks, which is between three and four hours a week averaged, with the load concentrated in weeks two, seven, and eight. That is a part-time commitment for one person, and it is the realistic shape of a programme that produces a measurable result rather than a plan.
8. What Each Phase Should Not Include
The discipline that makes this roadmap work is as much about exclusion as inclusion. Each phase has a specific list of activities that appear productive and belong to a later phase, or to no phase at all.
| Phase | Do not do | Why it is excluded |
|---|---|---|
| 1. Clear | Writing articles; commissioning a redesign; buying backlinks; building a Wikipedia entry | Each produces output that is unreadable or unverifiable until the blockers are gone, and the redesign is the most expensive way to reintroduce them |
| 2. Build | Outreach for directory listings; press pitches; paid promotion of the new pages | The listings would point at pages that are thin, and the press would cover a site whose structural work is not yet verified |
| 3. Extend | Another round of structural work; a second language version; a new page type | The phase exists to corroborate and measure. Adding scope here means the measurement week arrives without a measurement |
The phase one exclusion list is the one that requires the most discipline, because every item on it feels like progress. Writing an article produces a document. Commissioning a redesign produces meetings and a schedule. Buying backlinks produces a report from a vendor. None of them produces a citation if the crawler is blocked, and all of them consume the attention that the six hours of week 1 tests require.
The phase three exclusion list is easier to follow and more commonly violated. A team that has just finished phase two is usually energised and wants to add a second language or a new page type, and the measurement week gets pushed. The result is a hundred and thirty hours of work with no comparison, which is the single least defensible outcome available.
9. The Two Documents That Make Weeks 9-11 Easier
The external listing work in phase three is the part where manufacturers most often lose track, because it involves eight sources, several logins, and facts that can drift. Two documents prevent the drift.
| Document | Contents | When to create it |
|---|---|---|
| Canonical facts page | Legal name, addresses, phone, founding year, certificates with numbers and dates, core product figures | Week 3, as part of the phase one consistency work |
| Listing ledger | Each source, its listing URL, whoever holds the login, the last verified date, and any known discrepancy | Week 9, before the first verification, and kept current as each source is checked |
The login column is the one that prevents the most wasted time. A manufacturer that has changed marketing staff twice often cannot edit its own association directory entry, and the fix begins with a password reset rather than with a content decision. Noting the holder at the time of the first verification costs a minute and saves an afternoon.
Both documents live with the programme owner rather than in a shared drive, because the value is in their being current. A canonical facts page that nobody has opened since month three is a page that describes a company that no longer exists in that form, which is the exact failure it was created to prevent.
10. Five Stall Points and How to Avoid Them
| # | Stall point | When it happens | How to avoid it |
|---|---|---|---|
| 1 | Week 1 becomes a content brainstorm | The first meeting, because content feels like the work | Fix the first meeting agenda to the four week 1 tests; no topic discussion permitted |
| 2 | Week 2 waits for a developer | The developer has other priorities and the request is not scheduled | Book the developer time before the programme starts, or confirm that the platform can do it without one |
| 3 | Week 7 stalls on the Q&A page | Nobody can agree which ten questions to include | Use the ten most-asked questions in the sales inbox, unchanged. The debate is the stall, not the choice |
| 4 | Week 9 reveals a mismatch that requires a certificate correction | The certification body takes weeks to respond | Submit the correction request in week 3 rather than week 9, since the check is cheap and the response is slow |
| 5 | Week 12 produces no comparison | The baseline was never captured in week 4 | Make the baseline a phase one acceptance test, so the programme cannot proceed without it |
Stall point 4 is the one that is entirely preventable by moving a task earlier. The certificate register check costs an hour and the correction, if one is needed, takes the issuing body days or weeks. Discovering the need in week 9 means the programme finishes with a known mismatch, which is the worst possible outcome for a ninety-day effort.
11. How to Decide at Day 90 Whether to Continue
| Signal | Continue | Reconsider |
|---|---|---|
| Citation count | Any movement, even from 0 to 1 on a query | No movement at all, with all five phase one tests passing |
| Attributed enquiries | One or more attributed to an assistant | Zero, with the form field in place and the CRM convention followed |
| Branded search | Rising volume during the period | Flat or falling |
| Enquiry quality | The sales team reports more specific opening questions | No change in the quality of the first message |
| Effort spent | Close to the 130-hour estimate | Far above it, which suggests a scope problem rather than a channel problem |
The reconsider row is genuinely worth acting on rather than pushing through. If all five phase one tests passed and there is no movement after ninety days, the cause is usually that the category is not being asked about — which is a market question rather than a visibility question, and the right response is to redirect the effort rather than to repeat it.
12. FAQ: GEO Roadmap Questions Manufacturers Ask Most
10.1 Can I compress this into 30 days?
Phase one can be compressed if the site is technically simple and a developer is available immediately. Phases two and three cannot, because the citation lag is measured in weeks and the external listings take time to appear. A 30-day version produces the work without the evidence, which is the least useful outcome.
10.2 Do I need a dedicated person?
Not full-time. About 130 hours across twelve weeks is roughly three to four hours a week averaged, concentrated in three weeks. One person with a partial allocation is enough, provided that person has access to a developer for weeks two and four.
10.3 What if I cannot get developer time in week 2?
Then start at week 3 and postpone week 2. The facts and the canonical page can be done without development, and the rendering fix can follow. The one thing not to do is proceed to phase two while the specifications are still JavaScript-rendered, because phase two then produces content that cannot be read.
10.4 How many articles should I publish during the 90 days?
None in phase one, and none required in phases two and three. The roadmap is about structural and decision-stage pages, not about article volume. If a content programme is running in parallel, that is fine, but it should not be the basis of the visibility effort.
10.5 Is 90 days enough to see a result?
Enough to see a movement, not enough to see the full effect. Perplexity-like surfaces can reflect changes within weeks; Google-based surfaces can take several months. A 90-day programme is designed to produce the first evidence and the case for a second quarter, not to complete the work.
10.6 What is the single week that decides the programme?
Week 1. If the crawler is blocked or the specifications are rendered client-side, every subsequent week produces work that cannot be read. The four week 1 tests cost six hours and they determine whether the other 126 hours are worth spending.
10.7 What if my site is in multiple languages?
Run the roadmap on the primary language version first, and extend to the other languages in a second cycle. Doing five languages in parallel multiplies the work without multiplying the learning, and the phase one tests are the same tests in every language.
10.8 Should I hire an agency for this?
An agency is useful for the markup work and for the external listings, and it is not necessary for the content pages, which require your own knowledge of your process and your products. The most common outcome of a fully outsourced programme is a set of pages that read as though they were written about a factory rather than by one.
10.9 How do I keep the programme alive after day 90?
Two habits: a quarterly re-run of the phase one tests, because site changes reintroduce the blockers; and a monthly citation grid, because the trend is the argument. Both take under an hour each and they are what keeps the programme from decaying into a completed project.
10.10 Can Kangdi share what our 90 days looked like?
We can share the shape: which phase took longer than planned, which acceptance test we failed the first time, and what the citation count looked like at weeks 4, 8, and 12. Send us a note with "90-Day Roadmap" and we will describe our own sequence, including the parts we would sequence differently if we ran it again.
13. What the Second 90 Days Looks Like
The first ninety days produce a foundation and a first measurement. The second ninety days produce scale, and the work is different enough that it is worth planning before you arrive at it.
| Area | First 90 days | Second 90 days |
|---|---|---|
| Structural | Remove blockers, fix rendering, add markup | Quarterly re-test only, plus a check after any site change |
| Content | Four decision-stage page types | A second language version, and the Q&A page expanded to thirty entries |
| Corroboration | Two registers verified, two platforms aligned | Association membership live, one trade press mention pursued, a commercial data provider record claimed |
| Measurement | Baseline and a first comparison | A monthly grid and a quarterly enquiry-source review |
| Effort | Roughly 130 hours | Roughly 40-60 hours, because the structural work is done |
The effort drop is the useful part of this table. The first ninety days are front-loaded because the structural work cannot be avoided, and the second ninety days are largely editorial. A manufacturer that has completed the first cycle and found the effort unsustainable has usually been doing structural work that should have been finished once, rather than continuing to rediscover it.
The second-cycle content decision deserves a specific note. A second language version is the highest-value content addition, provided the seven localisation elements are done — a translated page that leaves prices in the wrong currency and dates in the wrong format damages credibility more than it adds reach. If localisation capacity is limited, it is better to localise one market properly than to translate four markets superficially.
14. About Kangdi Medical and How We Ran Our Own 90 Days
Henan Kangdi Medical Devices Co., Ltd has manufactured transdermal patches since 1989 at a GMP-certified facility in Zhoukou, Henan, China, and we ship to 200+ brand buyers and distributors across 40+ countries. We ran this roadmap on our own site, and we would rather share what actually happened than describe a plan that was never tested.
Three steps to get the roadmap applied to your site. (1) Email us with "90-Day Roadmap" as the subject line and include your domain and the three pages that matter most. (2) We will run the five phase one tests and tell you how many already pass, which is the fastest way to know where your twelve weeks would actually be spent. (3) We will send a version of the calendar adjusted to your starting position, with the weeks that can be compressed and the ones that cannot.
- Step 1: Email export@kangdimedical.com with "90-Day Roadmap" as the subject line, including domain and three key pages.
- Step 2: Receive the five phase one test results within 3 business days.
- Step 3: Get the adjusted twelve-week calendar, with the compressible weeks marked.
Tell us your domain and the three pages that matter most, and we will send the phase one test results and the adjusted calendar within 3 business days.