You fixed the right things first. One big job is still open.
Since the first audit, your team fixed the sitemap, added structured data about IDA to every page, gave every page its own title and wrote real descriptions for every resource. Those fixes already show. IDA's score for AI answers went up from 58 to 63, and real visitors now get a fast site.
The open job is the old website. Search engines and archives know 7,809 old IDA addresses, and 397 of the 400 we checked now show "Not Found". Years of links, including links from Wikipedia, still point to them. This page shows what you fixed, how to bring that value back, the next ten steps, and the larger work we can do for you.
Your scores today
We use two separate scores. One measures how ready the site is to be quoted in AI answers from ChatGPT, Gemini, Perplexity and Google's AI Overviews. The other measures classic search health. A third number stands alone, because it is the biggest problem on this page.
AI answers
Same method as the first audit, run again on 23 September
Was 58 on 15 September. Up 5 points. The magenta segment is the gain.
Today, by category
The gain came from the structured data, the sitemap and the single homepage heading. Brand authority did not move, because Wikipedia, Wikidata and Reddit have not changed.
Search health
First time measured, seven weighted categories
A solid base with clear gaps.
Magenta marks a category under 50.
Link equity preservedThe value of the links that point at the old site's 7,809 addresses. None of those addresses redirect to the new site, so almost none of that value reaches it. See priority one.
What you fixed
The first six items below were on the first audit's list. Each was the right fix, and several are already paying off. The last two are good news we measured this time.
Good technical SEO is what AI tools use to find and trust a website. These fixes were the right place to start.
The XML sitemap is live
The sitemap lists 52 pages and robots.txt points to it. Search engines and AI crawlers now get a complete list of your pages.
Before: no sitemap at all.Structured data about IDA is on every page, 52 of 52
This is a small block of code that tells Google and AI tools who IDA is. It links to IDA's Wikipedia article and Wikidata entry (Q6050174), with the founding year and both offices. This single change moved AI platform readiness the most: ChatGPT readiness went from 58 to 73, and Gemini from 42 to 63.
Before: none.Every page has its own title
Google and AI tools can now tell your pages apart in results.
Before: 22 of 40 pages were titled just "IDA".Resource pages have their own accurate descriptions, 21 of 21
Each resource now describes itself in search results.
Before: one copy-pasted CRPD sentence on all of them.The homepage has one main heading
The homepage now tells search engines one clear thing about what IDA is.
Before: two.The facts now agree
The ASEAN member count is 10 everywhere on /our-network. The events listing now shows COSP19 in June, in New York.
Before: 10 in one place and 11 in another. COSP19 was listed as August, Geneva.Real visitors get a fast site, and it keeps getting faster
Largest Contentful Paint (how long the main content takes to appear) went from 2.9 seconds to 1.8 seconds over five weeks, measured on real Chrome visitors. Interaction speed is 68 ms and layout shift is 0.02. All three Core Web Vitals are good. See the speed chart.
All major AI crawlers can read the site
GPTBot, OAI-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended and the other major AI crawlers all get through. IDA's Domain Rating is 71, level with Humanity & Inclusion (72) and the World Federation of the Deaf (72).
Priority one
The old website's pages all lead to "Not Found"
The Wayback Machine holds 7,809 old IDA addresses from 2013 to 2026: 6,716 pages and 1,093 PDFs. We checked 400 of them live on 23 September, starting with the most visited. 397 show "Not Found". None redirect. The other 3 (/, /resources and /events) work only because the new site happens to use the same short address.
Why it matters
IDA's Domain Rating of 71 was built by years of links to those old pages. Every one of those links now ends on a "Not Found" page, so that authority is not passed to the new site. Google still shows many of these dead addresses in its results today. AI tools that check a source hit the same dead end, and people who click give up.
Domain Rating, IDA and its neighbours
Ahrefs, 0 to 100. Higher means more and stronger links.
IDA's 71 was built while the old site's addresses were live. Redirects are how that strength reaches the new pages.
Proof from real places
- IDA's own English Wikipedia article cites /secretariat. Click it: "Not Found".
- The WeThe15 Wikipedia article cites /ipc-ida. "Not Found".
- The IDDC Wikipedia article cites /ida-anniversary. "Not Found".
- Google searches for "International Disability Alliance executive director" and "IDA members" today show dead IDA addresses such as /chair, /secretariat, /about, /content/ida-members and /all-countries.
The shape of the old site, next to the new one
Old addresses by type, from the Wayback Machine. Each group has a sample dead link.
2,563 of the old addresses were Spanish pages under /es/, spread across these groups. Most old pages will fold into one new page, because the new site is much smaller.
The 20 old addresses to fix first
Chosen for who still links to them and whether Google still shows them. Click any old address to see the "Not Found" page for yourself. The suggested new home is the closest page on today's site.
| # | Old address (shows "Not Found") | Why it matters | Suggested new home |
|---|---|---|---|
| 1 | /secretariat | Cited by IDA's Wikipedia article | /our-team |
| 2 | /ipc-ida | Cited by the WeThe15 Wikipedia article | /new-releases |
| 3 | /ida-anniversary | Cited by the IDDC Wikipedia article | /new-releases |
| 4 | /CRPD | Still in Google results for "CRPD" | /resources |
| 5 | /about | Still in Google results | /our-network |
| 6 | /content/ida-board-and-secretariat | Still in Google results | /our-team |
| 7 | /content/ida-members | Still in Google results for "IDA members" | /our-network |
| 8 | /content/legal-and-financials | Still in Google results | /governance |
| 9 | /generalassembly | Still in Google results | /governance |
| 10 | /UnderstandingtheCRPDHandbook | Still in Google results, a long-lived resource | /resources |
| 11 | /COSP | Still in Google results for "COSP" | /events |
| 12 | /content/crpd-committee | Still in Google results | /resources |
| 13 | /CRPD-advocacy | Still in Google results | /global-influence-and-research-center |
| 14 | /content/history | Still in Google results | /our-network |
| 15 | /content/our-principles | Still in Google results | /our-strategy |
| 16 | /all-countries | Still in Google results | /our-network |
| 17 | /crpd-implementation-guidelines | Still in Google results | /resources |
| 18 | /content/associate-members | Still in Google results, and recent | /our-network |
| 19 | /sites/default/files/ida_articles_of_association.pdf | A governance document others link to | /governance |
| 20 | /sites/default/files/ida_global_survey_complete_report_final.pdf | Still in Google results | /resources |
Site settings, then Publishing, then 301 redirects. Webflow supports bulk import and pattern rules. It still needs care:
We have the full list of 7,809 old URLs, with a suggested destination for each one. This is the first job we would do.
One question for IDA
The old site had 2,563 Spanish pages, such as /es/resources. The new site has none. Does IDA plan to bring Spanish back? The answer decides where those addresses should point.
Top 10 next steps
In order of value. Each card says who is best placed to do it.
-
Redirect the old URLs
Send each of the 7,809 old addresses to the closest new page, so links from Wikipedia, Google and partner websites land somewhere useful again. See priority one.
-
Fix the broken link on three pages
The breadcrumb on /governance, /our-network and /our-strategy points to /our-purpose-and-commitment, which shows "Not Found". It is one shared component, so one fix repairs all three pages.
-
Remove the stray period at the end of 35 page titles
The CMS title templates end with "International Disability Alliance." and that full stop shows in every Google result for those pages. One edit per collection template fixes all 35.
-
Give every release, resource and event a short, quotable summary
35 CMS items today. A summary that states the facts in two or three sentences is what AI tools quote. See B below.
-
Add the missing structured data
News data on the 7 releases, event data on the 7 events, person data for the team and board, and breadcrumb data on every page. Also fix the resource data, which uses an incomplete page address on all 21 resource pages and fails validation. See C below.
-
Link related pages to each other
Two program pages, /alliance-and-opd-strengthening and /inclusive-solutions-center, have no links in their content at all. See D below.
-
Correct the stale facts in Wikipedia and Wikidata
Wikipedia still names the previous Executive Director. We prepare the sourced request so an independent editor can make the change. See E below.
-
Bring /impact up to the standard of the rest of the site
It still has two main headings ("IMPACT" and "Enabling Good Lives"), a 6-character title ("Impact") and a 19-character description. And 14 of the 15 people in its story photos have no alt text, so screen readers and AI tools cannot tell who is pictured.
-
Fix the four known content errors
Nikola's story on /impact says Croatia in the heading and Serbia in the body. The privacy policy repeats "We may also collect information that you choose to provide, for example when you:" three times with no list after it, has joined words ("someactivities", "anapplication", "notresponsible"), and promises an update date it does not show. The COSP19 page says 8 June in one place and 9 June in another.
-
Add a source line under the homepage numbers
The homepage shows 430M lives impacted, +21K OPD-led interventions, +2,500 laws, +2,400 resources and 1.3 billion. None has a source or a year. AI tools and journalists repeat numbers they can check.
Valuable work to be done
Some of the most useful work left touches every CMS item, or thousands of old addresses, at once. In Webflow that is slow to do by hand, one item at a time. This is where we can help. Each item shows the count and one real example.
The redirect map
7,809 old addressesA destination for every old address, checked against the new site, in the order Webflow needs. The top 20 are in priority one. This is the job that brings back the most value for the least change to the site.
Quotable summaries for AI answers
35 releases, resources and eventsAI tools quote short passages that state facts plainly: who, what, where, when, and a number. Today no passage on the site reaches the "citation-ready" level. Every page we tested averages a C, D or F on the citability scale, and the /new-releases listing has no passage long enough to score at all.
The fix is a summary field on each release, resource and event. It shows under the headline, and it is reused as the page description and in the structured data. Here is one example, from the Tonga release:
"The International Disability Alliance (IDA) warmly welcomes the Kingdom of Tonga's ratification of the United Nations Convention on the Rights of Persons with Disabilities (CRPD), a historic milestone for persons with disabilities in Tonga and throughout the Pacific."
"Tonga is a Pacific Kingdom that ratified the UN Convention on the Rights of Persons with Disabilities (CRPD) on 4 September 2026. The treaty enters into force for Tonga on 4 October 2026, one month after ratification. The International Disability Alliance (IDA) says the milestone follows years of advocacy by Tongan organisations of persons with disabilities, led by the Lavame'a Ta'e'iloa Disabled People Association (LATA) and the Pacific Disability Forum, with IDA President Dr Nawaf Kabbara calling it a new chapter for Tongatapu, Vava'u, Ha'apai, 'Eua and the Niuas."
Two more places where the facts are there but AI tools cannot use them:
- The homepage paragraph about what IDA is has the right facts, but the numbers sit in separate counter widgets. AI tools cannot read them as one statement. One answer paragraph with the numbers inside it fixes that.
- All 21 resource pages are short teasers in front of a PDF. AI tools that do not open PDFs never see IDA's best material. The COP30 climate briefing PDF names a climate finance goal of 300 billion dollars a year and five specific asks. The web page for that briefing mentions neither.
Structured data on every template
4 templates, plus 21 resource pages to correctStructured data is how a page tells Google and AI tools what it is, in a form machines read without guessing. IDA now has it for the organisation. The next layer describes each page:
- News data on releases, with the publish date and the spokesperson. It says "this is news from this date". Today the releases have no machine-readable publish date at all.
- Event data on events, with dates and place. It says "this is an event on these dates at the UN in New York", as for COSP19.
- Person data for about 29 team members and the board on /our-team and /governance. It says "this person leads IDA".
- Breadcrumb data on every page, so search results show where a page sits on the site.
- A correction on all 21 resource pages, whose page address in the data is incomplete and fails validation.
Internal links between related pages
30 new links found, 7 CMS collectionsLinks inside the content tell Google and AI tools which pages belong together. We counted 1,239 internal links across the 52 pages.
Where the site's links sit
Most of the 274 in-content links are listing cards. Very few link one story to a related one.
We found 30 specific links, using words already on the pages. Four samples:
- On /impact, Chinedu's story ("Nigeria") should link to the African Disability Forum on /our-network.
- On /impact, Litia's story ("Fiji") should link to the Pacific Disability Forum on /our-network.
- In the Tonga release, "Dr Nawaf Kabbara" should link to his entry on /governance.
- The COP30 climate briefing should link to the /events/cop30 event page.
Also: /alliance-and-opd-strengthening and /inclusive-solutions-center have zero links in their content. /events shows 6 of 7 events. The seventh, the Second World Summit for Social Development, sits on a second page and no other content links to it.
Behind this: 7 CMS collections (releases 7, resources 21, events 7, team about 29, networks 14, impact stories about 16, board about 43) have no topic tags and no links to each other. Tags and "related" fields would let every new item link itself.
Wikipedia and Wikidata
1 article, 1 data entryChatGPT and Gemini lean on Wikipedia. Your website is now correct, but AI answers will keep repeating the old facts until Wikipedia changes.
| Fact | What Wikipedia and AI answers say | What IDA's website says |
|---|---|---|
| Executive Director | Vladimir Cuk | José María Viera, on /our-team |
| Size of the alliance | An AI answer today: "over 1,100 organisations… eight global and six regional networks", repeated from Wikipedia-era sources | 5,500 organisations, 14 networks, 182 countries, on /our-network |
| Last edited | Wikipedia: 14 December 2025. Wikidata Q6050174: 24 October 2025 | The site was last published on 23 September 2026 |
| Wikidata details | No founding date, no headquarters, no director, no LinkedIn or YouTube | Your structured data already states the founding year and both offices |
| Spanish | No Spanish Wikipedia article exists | No Spanish pages on the new site |
IDA staff should not edit the Wikipedia article themselves. Wikipedia's conflict-of-interest rules treat that as a problem even when the facts are right. The correct route is a sourced edit request on the article's Talk page, with independent sources for each fact. We prepare that request. Wikidata can be updated directly.
How ready each AI platform is for IDA today
Readiness score out of 100, 23 September.
Four of five platforms moved up sharply after the structured data and sitemap fixes. The Wikipedia work is the next big step for ChatGPT and Gemini.
An HTML sitemap page and an archive of the old site's content
52 pages today, years of old materialA simple page that lists every page on the site, grouped by topic, and an archive that keeps the best of the old site's content online. Years of material stay findable for search engines and AI tools, instead of 52 pages carrying the whole brand.
I promised Jarrod an HTML sitemap page and an archive for the old content. Both are in this plan.
Alt text and descriptions in bulk
14 portraits, 12 pagesAlt text for the 14 impact-story portraits on /impact. Only one of the 15 people pictured is described today. For a disability-rights organisation this is a credibility point as well as an accessibility one.
Descriptions for the 12 pages without one, most of them hub pages such as /resources, /events, /new-releases, /our-network and /governance. Without one, Google writes its own snippet for exactly the pages people click first.
We can do this work for IDA
Everything in A to G is work our team can take on, with your web team and staff approving each step. The redirect map is the first job, because every week of "Not Found" pages costs IDA links, search results and trust.
Write to Nick at nick@nickgray.net or call +1 917 635 9967.
Speed
Real visitors get a fast site, and it is getting faster. This is the measure Google uses.
How fast the main content appears for real visitors
Largest Contentful Paint, 75th percentile, Chrome UX Report, whole site. Each point is a 28-day window. The last five weeks are shaded.
Lab tests are slower, and that is normal
Google PageSpeed Insights also runs a lab test on a slow simulated phone. Those scores are middling, mainly because of animation code and Webflow's JavaScript files. Because real visitors already see good speed, this is a lower priority than everything above.
| Page | Mobile score | Mobile LCP (lab) | Desktop score |
|---|---|---|---|
| Homepage | 74 | 4.4 s | 83 |
| /impact | 69 | 4.8 s | 98 |
| A release: new UN guidelines on intersectional discrimination | 76 | 5.0 s | 100 |
| /resources/annual-report-2025 | 79 | 4.5 s | 99 |
The homepage takes 14.3 seconds in the lab before it responds to taps, and its hero image is set to load late although it is the first thing people see. Both are in everything else.
Everything else
Smaller items, medium and low priority. Open any one for the detail.
No canonical tags on any page Medium
0 of 52 pages have a canonical tag, the line that tells search engines which address is the official one for a page. Without it, versions with and without "www" or a trailing slash can compete with each other as the site grows.
The 2025-2030 strategy has no real web page Medium
/our-strategy is 165 words in front of a PDF. IDA's strategy for 2025 to 2030 deserves a real summary on the page, so search engines and AI tools can read and quote it.
No named author on releases Medium
Every release is signed "IDA Media". A named spokesperson, linked to their bio on /our-team or /governance, gives Google and AI tools a real person to trust.
The homepage hero image loads late Low
The large photo at the top of the homepage is set to "lazy" loading, which delays the first thing people see. Remove lazy loading on that one image. The homepage mobile lab score is 74, with a lab LCP of 4.4 s and 14.3 s before it responds to taps, caused by animation code and Webflow's JavaScript files. Real visitors see good speed, so this is a lower priority.
Social sharing tags are incomplete Low
The og:url and og:site_name tags are missing on all 52 pages. The twitter:image tag is missing on 17 pages, including the homepage, so some shares show no picture.
A placeholder page is in the sitemap Low
/coming-soon is listed in the sitemap, but no page links to it. Remove it from the sitemap.
No llms.txt file Low
/llms.txt shows "Not Found". This is an optional file that gives AI tools a short guide to a site. Google ignores it. It is cheap to add.
No favicon file at the standard address Low
/favicon.ico shows "Not Found". The icon tags on the pages exist, so browsers are fine. Some tools still look for this file.
Two AI crawlers are blocked Low
robots.txt blocks Amazonbot (Alexa) and YouBot (You.com). Their audiences are small for an NGO, so this is fine if it was a deliberate choice.
Only one security header is set Low
Only HSTS is set. The other standard security headers can be added at Cloudflare in one place.
The short address takes two redirects Low
http://internationaldisabilityalliance.org takes two redirects to reach the final address. One is enough.
Generic link wording Low
There are 40 generic "See all…" links and 23 "Download" links that do not name the document, for example on the resource pages. "Download the CRPD at 20 report (PDF)" tells people, screen readers and AI tools much more than "Download".
How we did this
We ran two independent audit methods on 23 September 2026. One is built for search engines: technical health, on-page, structured data, content, images and speed. The other is built for AI answers (GEO): citability, AI crawler access, brand and entity signals, and readiness for each AI platform. We checked all 52 pages in the sitemap. The AI answers score uses the same method as the first audit.
Sources
- 7,809 old IDA addresses from the Wayback Machine, 400 of them checked live on 23 September.
- Google PageSpeed Insights for lab tests, and the Chrome UX Report for real-visitor speed.
- Ahrefs Domain Rating for IDA and six comparable organisations.
- Live Google searches for IDA's name, leadership, members and core topics.
- IDA's Wikipedia article and Wikidata entry.
How the search health score is built
| Category | Weight | Score |
|---|---|---|
| Content | 23% | 42 |
| Technical | 22% | 66 |
| On-page | 20% | 62 |
| Schema | 10% | 38 |
| Performance | 10% | 71 |
| AI readiness | 10% | 60 |
| Images | 5% | 47 |
| Weighted total | 100% | 56 |
What we could not see
We did not have access to IDA's Google Search Console, so index coverage and search traffic figures are not included.