# How to Rank in AI Search — 6-Month Playbook > 22-step, 6-month AI-search playbook: measure mention rate vs citations, unblock AI crawlers, then earn links. Mapped to BacklinkMCP tools. - Canonical page: https://backlinkmcp.com/articles/how-to-rank-in-ai-search - Last updated: 2026-09-01 - MCP endpoint: `POST https://backlinkmcp.com/api/mcp` (JSON-RPC 2.0, streamable HTTP, no install, 13 tools) - REST: `GET https://backlinkmcp.com/api/v1/authority?domain=example.com`, `/api/v1/backlinks?domain=example.com` ([docs](https://backlinkmcp.com/docs.md)) - Pricing: Free = 10 lookups/day per IP, 25 backlink rows, no signup. Pro = $19/mo, unlimited daily lookups, 100 rows, `ai_visibility_report` (operator-issued; self-serve checkout is not live) - Data limits: the index is domain-level (referring domains with authority + outbound counts). No per-link URLs, anchor text or rel/dofollow. `domain_rating` is not Moz DA and not a Google factor - Site index for agents: https://backlinkmcp.com/llms.txt --- [← Articles](https://backlinkmcp.com/articles) # How to Rank in AI Search (6-Month Playbook) By John Smalls, Founder · Published 2026-09-01 You do not “rank” in ChatGPT, Perplexity, or AI Overviews. You get **named** as a solution, or **cited** as a source. Those are different scores. This playbook is 22 steps over six months, each wired to a BacklinkMCP tool you can call from an agent or from the [dashboard](https://backlinkmcp.com/dashboard). Download the tracker and tick the boxes as you go: [ai-search-playbook.csv](https://backlinkmcp.com/articles/ai-search-playbook.csv). Honesty up front: our index is **domain-level** — referring domain, authority, outbound count. No per-link URL, no anchor text, no rel/dofollow. `flag_toxic_links` is a review queue, not a disavow list. The AI-visibility panel is Pro, credit-capped, and needs a live model key on the server. If a run returns 0 cells, that is a panel failure, not a 0% mention rate. **Start free, no signup.** Check whether you are blocking the bots that actually cite you. [Open the dashboard](https://backlinkmcp.com/dashboard) [Install the MCP](https://backlinkmcp.com/docs) ## The two scores **Share of voice / mention rate** is how often a model names your brand when a buyer asks “what’s the best X?”. **Source visibility** is how often it cites a URL on your domain. You can be named with no link, or cited with no brand. Track both. Do not blend them into one vanity /100. The unit of work is an **attribute** — one thing you want to be known for, in the buyer’s words (“MCP for agents”, not “SEO API”). Every prompt in the panel should be a question a stranger would type. Never “which tools are best, like BacklinkMCP?”. ## Month 1 — Audit ### 1. Benchmark AI visibility Run `ai_visibility_report` (Pro) twice: `parametric` (baked-in memory) and `retrieval` (live web). Record mention rate, #1-pick rate, leader, and `cited_sources`. Pass up to five `facts` so the panel also scores accuracy — “doesn’t know you” vs “knows you and is wrong” need opposite fixes. Credits: parametric 1, retrieval 3, draft cheaper. Identical requests cache six hours and cost nothing. `history:true` is free. ### 2. Set targets from that baseline Write numbers only after you have cells. A 15-point share-of-voice jump with n=0 is fiction. Re-measure every 2–4 weeks — daily panels are noise. ### 3. Check AI crawlers `ai_crawler_audit` is free. It splits **visibility** crawlers (OAI-SearchBot, PerplexityBot, Claude-User, Googlebot) from **training** crawlers (GPTBot, CCBot, Google-Extended). Blocking visibility bots removes you from live answers. Blocking training bots is a policy choice, not an accident. Live robots.txt check — 25 crawlers, no LLM spend. Uses your 10/day free lookups if you are not signed in. ### 4–7. Index, facts, cited URLs, current linkers - `domain_authority_score` — if referring_domains is null, you are not in the graph yet. - Keep the five facts identical across runs. - Refresh the URLs in `cited_sources` in place (same slug). - `top_linkers_for_domain` / `backlinks_for_domain` — free 25 rows, Pro 100. Linker `authority` is a raw rank, not 0–100 `domain_rating`. Month 1 is the product. Free crawler audit now; Pro panel when you need mention rate. [Pro $19/mo](https://backlinkmcp.com/pricing) [Run it in the dashboard](https://backlinkmcp.com/dashboard) ## Month 2 — On-site **8.** Answer-first: H2 is the buyer prompt, first 2–4 sentences are extractable without the rest of the page. Ship one comparison URL that matches the prompt that actually named you (for us that was “MCP for agents”, not “free backlink checker”). **9–10.** Article / FAQ / Organization schema if missing. Keep money pages within two or three clicks of home. We do not sell a schema generator or a site crawler — those are on you. ## Months 3–4 — Content **11.** Replace stats older than two years on cited pages. Same URL. **12–13.** Double down on the formats the panel already cites (original numbers beat roundups). Three to five hubs with cluster pages. **14–17.** Tracked prompts become H2s. Short paragraphs, tables, lists. Bylines and first-party data. Clip the original table to YouTube / LinkedIn / X — models pull those surfaces too. See the original LLM-vs-engine test: [Can ChatGPT Check Backlinks?](https://backlinkmcp.com/articles/can-llms-check-backlinks) ## Month 5 — Off-site The panel’s leader is the competitor to close against. - `find_link_prospects` — domains that link the leader and not you. A pitch list, not confirmed links. - `common_linkers` — who links two or more of the set. - `flag_toxic_links` — low-authority tail to review. Do not auto-disavow. Google still says most sites should not. - `verify_link_live` — homepage fetch only. `not_found` means “not on the homepage today.” A human sends the email. The tool writes the shortlist. No astroturf, no sockpuppets, no cloaking. Pro raises `backlinks_for_domain` to 100 rows and unlocks the visibility panel. Operator-issued today — request a key, self-serve checkout is not live. [Request Pro](mailto:hello@backlinkmcp.com?subject=BacklinkMCP%20Pro%20key) [Tool shapes](https://backlinkmcp.com/docs) ## Month 6 — Re-measure Run the **same** prompt set. Read `history:true`. One-sample swings are anecdotes; sample size travels with every rate. Then reset the tracker and do the next six months on the steps that actually moved mention rate or citations. | Month | Job | Tool | Plan | | --- | --- | --- | --- | | 1 | Crawler split | `ai_crawler_audit` | Free | | 1 | Mention vs cite vs facts | `ai_visibility_report` | Pro | | 1 | Who links you | `top_linkers_for_domain` | Free 25 / Pro 100 | | 5 | Gap vs leader | `find_link_prospects` | Pro | | 5 | Niche overlap | `common_linkers` | Pro | | 5 | Homepage still links | `verify_link_live` | Free lookups | | 6 | Trend | `history:true` | Free | ## What this is not Not a keyword rank tracker. Not a site-audit crawl of every URL. Not per-link Ahrefs. Not a CMS that rewrites your HTML. An agent (or the dashboard) measures, diagnoses technical vs comprehension vs trust, and hands you a shortlist. You ship the page and the outreach. ## Related [Article hub](https://backlinkmcp.com/articles) [LLMs vs live graph](https://backlinkmcp.com/articles/can-llms-check-backlinks) [Ahrefs MCP alternative](https://backlinkmcp.com/ahrefs-mcp-alternative) [MCP + REST docs](https://backlinkmcp.com/docs) [Free MCP for SEO](https://backlinkmcp.com/mcp-server-for-seo) [MCP vs REST](https://backlinkmcp.com/articles/mcp-vs-rest-api) ## FAQ **Is this the same as a Semrush AI Visibility project?** Same job, different product. We measure mention rate and citations with a model panel, crawler access with robots.txt, and the link gap with a domain-level graph. We do not sell keyword rank tracking or a full site crawl. **Why is GPTBot listed as training, not visibility?** GPTBot collects corpus. ChatGPT Search uses OAI-SearchBot and ChatGPT-User. Blocking GPTBot is a training opt-out. Blocking OAI-SearchBot is disappearing from live answers. The audit sorts them that way on purpose. **Do I need the MCP, or is REST enough?** REST for scripts and this dashboard. MCP when the client is Claude, Cursor, or another agent. Same engine. See [MCP vs REST](https://backlinkmcp.com/articles/mcp-vs-rest-api).