AI visibility
How do you track AI search algorithm updates effectively?
Updated October 2026 · Manoj Surya
Quick answer
MarketHQ tracks AI search updates by doing what works without a changelog: ask the same fixed buyer questions in each AI assistant every week and record who is named and which sources are cited. An update shows up as many answers shifting at once, on one engine, in the same direction. One changed answer is usually noise. Compare engines to tell an update from your own content changes.
Key points
- AI assistants rarely publish ranking changelogs, so your own weekly baseline is the detector.
- A real update moves many questions on one engine in the same week; noise moves one or two at random.
- MarketHQ asks your tracked questions in ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek every week on paid plans.
- MarketHQ records the brands named and the source links in each answer, so you can see what the engine started favouring.
- Read vendor release notes too, but treat them as context, not as your measurement.
Founder of MarketHQ. Building market intelligence for companies that sell to developers.
Google publishes notes on many of its ranking updates. AI assistants mostly do not. When ChatGPT or Perplexity changes how it picks sources or brands, the only sign is that your answers look different. This guide shows how to detect those changes from your own data and tell them apart from noise.
Why are AI search updates harder to track than Google updates?
Because AI assistants change models, retrieval and source choices often and mostly without a public changelog, and there is no position number to watch.
With classic search you can watch a rank tracker and an industry volatility chart and line them up against an announced update. AI assistants work differently. A new model version, a change to how the assistant searches the web, or a new source partnership can each change which brands get named. You find out when your answers change.
There is also no single position. An answer either names you or it does not, and it may describe you well or badly. So the signal you track is a share across many questions, not a rank on one.
What baseline do you need before you can spot an update?
A fixed list of buyer questions, asked on the same day each week in each engine, with the same fields logged every time.
Without a baseline there is nothing to compare against. Write 10 to 30 questions a buyer asks before they know your name. Keep the wording fixed. Ask each in every engine you care about and log the result.
- Date, engine and exact question.
- Whether your brand was named, and where in the answer.
- Which other brands were named.
- Every source link the answer cited.
- A short note if the answer described you wrongly.
How do you tell a real update from normal variation?
Look for breadth and direction: many questions on one engine changing in the same week, in the same way, and staying changed the next week.
AI answers vary from run to run. If one question drops you one week and brings you back the next, that is noise. An update looks different. Your share on one engine falls or rises across most of your questions at once, the set of cited domains changes, and the new pattern holds for a second week.
Comparing engines is the most useful check. If ChatGPT shifts and Claude, Gemini and Perplexity do not, the cause is probably on ChatGPT's side. If every engine shifts together, look at your own site, a competitor launch, or a roundup that changed.
| Pattern | Likely cause | What to do |
|---|---|---|
| One question changes, then reverts | Normal variation | Nothing; keep watching |
| Many questions shift on one engine, and hold | A change at that engine | Compare cited sources before and after |
| All engines shift together | Something changed on the web | Check your pages, competitors and key roundups |
| New domains appear in citations | The engine's source choice moved | Get listed on the new sources |
Where do you find official news about AI search changes?
On each vendor's own release notes, help centre and blog, which cover product changes but rarely explain how sources or brands are chosen.
Follow the release notes and help pages of the assistants your buyers use. They announce new models, features and search behaviour. Use them to explain a shift you already measured, not to predict one. A changelog entry with no change in your answers is not worth reacting to.
How does MarketHQ track AI search changes for you?
It runs your fixed question list in six AI assistants every week on paid plans and keeps the history, so shifts are visible without a spreadsheet.
Your tracked questions go to ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek each week. Each answer records whether you were named, who was named instead, the answer text and the source links it included. Because the wording never changes, a week-on-week shift on one engine stands out.
Questions you lose feed the weekly gap analysis and action plan, alongside community, news and social mentions of you and your competitors. Lite is $29 a month for 10 questions; Startup is $99 for 60; Growth is $199 for 120.
What does a detected update look like?
Illustrative example: one engine's citations move from vendor pages to forums in a single week.
Illustrative example: a payroll software company tracks 20 questions. For two months it is named in about half of Perplexity's answers. Then, in one week, it drops to a quarter on Perplexity while the other engines hold steady. The cited links explain it: Perplexity now leans on two Reddit threads and a review site, and it cites fewer vendor pages. The company answers new questions in those communities with disclosed, useful replies and updates its review profile. Over the next weeks its share on Perplexity climbs back.
What are the limits of tracking AI search updates?
You only see changes on the questions and engines you track, and a tracker's answers can differ from a signed-in user's.
A small question list can miss a change that affects other topics. Trackers usually ask through an API, which can differ from the consumer app. MarketHQ does not track Google AI Overviews, Google AI Mode or Microsoft Copilot. Treat each week as one reading and act on trends that hold.
FAQ
- What tool tracks AI search algorithm updates?
- MarketHQ asks your fixed buyer questions in ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek every week and keeps the history, so a shift on one engine is visible in your own data. Plans start at $29 a month. Vendor release notes add context but rarely explain source choices.
- How often should I check for AI search changes?
- Weekly, on the same day. That is often enough to catch a shift within days of it happening and slow enough to avoid reacting to single-run noise. After you see a change, keep the same schedule and confirm it holds before you rework pages.
- How is this different from tracking Google algorithm updates?
- Google announces many core updates and you can watch a ranking position. AI assistants rarely announce source or ranking changes, and there is no position, so you measure share of answers that name you across a fixed question set.
- Can I use my SEO rank tracker for AI search?
- Partly. It still shows how your pages rank, and well-ranked pages are often cited. It does not show whether ChatGPT or Claude names you. MarketHQ adds that weekly check across six assistants next to your keyword tracking.
- How do I know a change is real and not random?
- Check breadth, direction and persistence. A real update moves many questions on one engine the same way and holds for a second week. One question flipping is usually normal variation.
Related
- Weekly AI visibility routine for founders — a 30-minute weekly habit
- ChatGPT vs Perplexity — why engines disagree
- AI citation tracking — watching the sources answers cite
- AI visibility use case — how MarketHQ runs the weekly check